TL;DR: Google rolled out AI Overviews to nearly a quarter of all search queries by July 2025 (24.61%), then quietly rolled them back to 15.69% by November—a 36% decrease nobody at Google bothered to announce. Semrush data tracking 10+ million keywords shows the peak happened in mid-summer, followed by a dramatic pullback. Meanwhile, Ahrefs analysis of 540,000 query pairs reveals AI Overviews and AI Mode cite the same URLs only 13.7% of the time despite reaching 86% semantic similarity. They’re completely separate systems using different sources. Google’s lead product manager Logan Kilpatrick confirmed in September 2025 that AI Mode will become the “default” search experience “soon.” The rollback isn’t a failure—it’s strategic repositioning. Google is quietly phasing out AI Overviews while preparing AI Mode for mass adoption. Publishers who optimized for AI Overviews are now facing a system that cites completely different sources.

Nobody at Google announced this.

In July 2025, AI Overviews appeared for 24.61% of all search queries tracked by Semrush.

By November 2025, that number had dropped to 15.69%.

That’s a 36% decrease in AI Overview visibility in just four months.

No blog post. No announcement. No explanation.

Google just quietly rolled back one of the biggest changes to search results in a decade, and hoped nobody would notice.

But Semrush noticed. They tracked 10+ million keywords from January through November 2025.

And the data tells a story Google doesn’t want to tell.

The Rise and Fall Nobody Saw Coming

Here’s the complete trajectory:

January 2025: 6.49% of queries triggered AI Overviews July 2025: 24.61% of queries triggered AI Overviews (peak) November 2025: 15.69% of queries triggered AI Overviews

The rise was aggressive. January to July saw a 279% increase in AI Overview visibility.

Google was flooding search results with AI-generated summaries. Nearly one in four searches showed an AI answer at the top of the page.

Then something changed.

Between July and November, AI Overviews dropped from appearing on 24.61% of queries to 15.69%—losing more than a third of their presence.

The decline was just as aggressive as the rise. But this time, Google said nothing.

What Google Isn’t Saying

When AI Overviews launched globally in May 2024, Google positioned them as the future of search.

Liz Reid, head of Google Search, called them “a new era” for how people find information.

The company expanded AI Overviews to over 120 countries. They integrated ads into AI Overview results. They made the feature central to every major Search announcement.

Then they started pulling it back.

And when Semrush published data showing the rollback in mid-December 2025, Google’s response was silence.

No statement explaining the decrease. No guidance about whether this is temporary or permanent. No clarity about what changed.

Just a 36% reduction in visibility that affected hundreds of millions of searches per day.

The Pattern Hidden in the Data

The Semrush study reveals something even more interesting than the overall decline.

Q2 2025 showed an explosion of AI Overviews across every industry vertical.

Then Q3 saw a pullback—more than half of all sectors experienced decreases in keywords triggering AI Overviews.

Q4 brought increases again, but not to Q2 levels.

This isn’t random volatility. It’s testing and correction.

Google pushed AI Overviews hard in mid-2025, measured the results, then made strategic adjustments based on what they found.

What did they find that made them reduce AI Overview visibility by 36%?

Why Commercial Queries Got Cut

Early 2025 AI Overviews were almost entirely informational. In January, 91.3% of queries triggering AI Overviews were purely informational.

By October, that share dropped to 57.1%.

Commercial queries increased to 18.57%. Transactional queries reached 13.94%. Navigational queries (brand searches) jumped from 0.74% to 10.33%.

Google was expanding AI Overviews into money-making queries—the searches that drive ad revenue.

Then they pulled back.

Here’s why: ads alongside AI Overviews rose from 3% in January to roughly 40% by November. But 95% of keywords triggering AI Overviews either displayed no paid ads or had minimal commercial value.

Google’s $200+ billion advertising business depends on users clicking ads on commercial queries.

When AI Overviews started appearing on commercial searches, they displaced those ads.

The math became simple: keep AI Overviews on informational queries where ad revenue is low, reduce them on commercial queries where every click matters.

The Zero-Click Myth That Won’t Die

Critics predicted AI Overviews would destroy organic traffic by creating zero-click searches.

The data tells a different story.

When Semrush analyzed the same keywords before and after AI Overviews appeared, zero-click rates actually decreased—from 33.75% to 31.53%.

Overall zero-click rates for keywords with AI Overviews have steadily declined since January 2025.

AI Overviews don’t automatically reduce clicks. They often appear on queries that were already unlikely to generate clicks.

But here’s where it gets complicated: click-through rates for top-ranking pages dropped 34.5% when AI Overviews were present. Some publishers reported 50-90% lower CTRs on queries where AI summaries appeared.

Seer Interactive’s study of 3,119 informational queries across 42 organizations found organic CTR dropped 61% year-over-year (June 2024 to September 2025) on queries where AI Overviews were present.

The zero-click myth is wrong. But the click redistribution reality is devastating for publishers.

Enter AI Mode: The Real Replacement

While everyone focused on AI Overviews declining, Google was quietly building something else.

AI Mode launched in May 2025 as an “experimental” search experience.

By July, usage had increased 4x—from 0.25% in May to over 1% of all searches.

In September 2025, Logan Kilpatrick, Google’s lead product manager for AI, responded to a question about AI Mode becoming the default experience.

His answer: “Soon.”

Not “maybe.” Not “we’re considering it.”

“Soon.”

Liz Reid had already signaled this direction in May 2025, calling AI Mode “the future of Google Search.”

Now Google’s own product manager confirmed the timeline: AI Mode will replace traditional search as the default experience.

Two Systems, Completely Different Sources

Here’s where the story gets interesting.

AI Overviews and AI Mode are both Google products. Both use generative AI. Both appear in Google Search.

Most people assumed they were variations of the same system.

Ahrefs analyzed 540,000 query pairs comparing AI Mode and AI Overview responses.

The results: AI Mode and AI Overviews cite the same URLs only 13.7% of the time.

When restricting the comparison to just the top three citations, overlap increased slightly to 16.3%.

That still means 87% of citations are completely different.

SE Ranking found similar results: 10.7% URL overlap and 16% domain overlap between the two systems.

Despite this massive citation divergence, the two systems reach semantically similar conclusions 86% of the time.

They agree on what to say while fundamentally disagreeing on where they found it.

Why the Citation Difference Matters

Publishers who spent months optimizing for AI Overviews are about to face a brutal reality.

Visibility in AI Overviews does not translate to visibility in AI Mode.

Only 13.7% of citations overlap. That means 86.3% of the time, being cited in an AI Overview does nothing for your AI Mode visibility.

And if AI Mode becomes the default experience “soon,” all the optimization work done for AI Overviews becomes partially obsolete.

Different systems. Different sources. Different strategies required.

The Platforms That Win in Each System

Ahrefs’ research reveals distinct source preferences:

AI Mode favors:

• Wikipedia (28.9% of responses vs 18.1% for AI Overviews)

• Quora (3.5x more citations than AI Overviews)

• Encyclopedic and detailed sources

• Longer, more comprehensive content

AI Overviews favor:

• YouTube (most frequently cited domain)

• Videos and core pages (nearly 2x more than AI Mode)

• Reddit (similar rates in both)

• Article-format content

The citation patterns aren’t just different—they’re optimized for different user intents.

AI Overviews aim for quick, digestible answers. AI Mode builds comprehensive explorations.

One is a summary. The other is a research assistant.

The Usage Patterns That Explain Everything

AI Mode sessions behave completely differently from AI Overviews:

Time spent:

• AI Mode: 49 seconds average

• AI Overviews: 21 seconds average

Zero-click rates:

• AI Mode: 93% of searches end without a click

• AI Overviews: 43% result in zero clicks

Session behavior:

• AI Mode: 75% of sessions never leave the pane

• AI Overviews: Users frequently click through to sources

AI Mode is designed to keep users in Google. AI Overviews still drive some external traffic.

That difference explains why Google is phasing out AI Overviews while positioning AI Mode as the default.

The Three Industries Google Abandoned

The Semrush data shows which industries lost the most AI Overview coverage between March and November 2025:

Health: Significant decrease in AI Overview share People & Society: Greatest decrease of any vertical Law: Major reduction in coverage

These are all YMYL (Your Money, Your Life) topics where inaccurate AI summaries could cause real harm.

Google aggressively expanded AI Overviews into these areas in Q2, then pulled back in Q3 after seeing the results.

Meanwhile, Food & Drink saw the fastest growth (+7.25% since March) and is now the ninth-most impacted industry.

The pattern suggests Google is being selective about where AI summaries add value versus where they create risk.

What Publishers Are Actually Seeing

The Professional Publishers Association in the UK documented the real-world impact:

• Click-through rates falling 10-25% year-over-year despite stable rankings

• Lifestyle publisher: CTR dropped from 5.1% to 0.6%

• Automotive publisher: CTR fell from 2.75% to 1.71%

Digital Content Next found AI Overviews linked to a 25% drop in referral traffic in August 2025.

Some publishers reported 50-90% lower CTRs when AI summaries appeared.

These aren’t marginal changes. They’re business-threatening declines.

And they happened while Google was quietly reducing AI Overview visibility by 36%.

The Ad Revenue Protection Strategy

Here’s what the data reveals about Google’s priorities:

January 2025:

• Ads alongside AI Overviews: ~3%

• AI Overviews mostly on informational queries: 91.3%

November 2025:

• Ads alongside AI Overviews: ~40%

• AI Overviews on informational queries: down to 57.1%

Google tried expanding AI Overviews to commercial queries. When that threatened ad revenue, they rolled back the feature and started monetizing it more aggressively where it remained.

The 36% decrease in AI Overview visibility happened alongside a 13x increase in ad integration (from 3% to 40%).

Google is protecting the $200+ billion advertising business while still appearing innovative with AI features.

Why AI Mode is Different (and Worse for Publishers)

AI Mode represents a fundamental shift from AI Overviews:

Response length:

• AI Mode: ~300 words average

• AI Overviews: ~50 words average

Citations per response:

• AI Mode: 12.6 links average

• AI Overviews: 13.3 links average (slightly more)

Zero-click rates:

• AI Mode: 93% (more than 2x AI Overviews)

• AI Overviews: 43%

User behavior:

• AI Mode: Users spend 49 seconds, rarely click out

• AI Overviews: Users spend 21 seconds, frequently explore sources

AI Mode citations are displayed in a sidebar with ~7 unique domains. Only 7% of responses include links below the main answer.

Even when you’re cited, users aren’t clicking. 93% of AI Mode sessions end without visiting any external site.

The Timeline of What’s Actually Happening

May 2024: AI Overviews launch globally January 2025: 6.49% query coverage May 2025: AI Mode launches; Liz Reid calls it “the future of Google Search” July 2025: AI Overviews peak at 24.61% August 2025: First major reports of publisher traffic declines September 2025: Logan Kilpatrick says AI Mode will be default “soon” November 2025: AI Overviews drop to 15.69% (36% decline from peak) December 2025: Semrush publishes data; Google says nothing

The pattern is clear: Google tested AI Overviews aggressively, measured the impact on ad revenue and publisher traffic, then began strategic rollback while positioning AI Mode as the replacement.

What Google Gained from the Test

The July peak wasn’t a failure. It was data collection.

Google needed to understand:

• Which query types work well with AI summaries

• How users interact with AI-generated content at scale

• Where AI features threaten core advertising revenue

• What level of publisher backlash to expect

They got their answers. Now they’re implementing the strategy.

Reduce AI Overviews on commercial queries. Increase ad integration where they remain. Position AI Mode—which keeps 93% of users inside Google—as the future default experience.

The SEO Industry Caught Flat-Footed

While everyone optimized for AI Overviews, Google was building a completely different system.

Only 13.7% citation overlap means strategies optimized for AI Overviews have limited transfer to AI Mode.

Different source preferences. Different content formats. Different user intents.

The industry spent 2025 figuring out how to get cited in AI Overviews.

That work might become partially obsolete when AI Mode becomes the default.

What “Soon” Actually Means

Logan Kilpatrick’s September 2025 statement that AI Mode will be default “soon” came with no specific timeline.

Google VP Robby Stein cautioned people not to “read too much into” the statement.

But the evidence suggests “soon” means 2026:

• AI Mode usage already increased 4x (from 0.25% to 1%+)

• google.com/ai now leads directly to AI Mode (launched in September)

• AI Mode rolled out to 180 countries

• 75 million users already adopted the feature

Google is preparing infrastructure for mass adoption. The default switch is implementation, not experimentation.

The Three Strategies That Still Work

Despite the chaos, three approaches remain effective:

1. Optimize for Both Systems Separately

AI Overviews favor quick-answer content. AI Mode favors comprehensive exploration.

Create both. Stop assuming one strategy covers both systems.

2. Focus on Brand Mentions Over Citations

AI Mode includes 2.5x more people and brand entities than AI Overviews.

If your brand appears in AI Overviews, there’s a 61% chance it’ll also show in AI Mode.

Brand recognition matters more than individual page citations.

3. Build for Discovery Outside Google

93% of AI Mode sessions end without clicks. Publishers can’t depend on Google sending traffic.

Reddit, YouTube, and community platforms matter more when Google’s AI keeps users inside the search page.

Why This Matters for Answer Engine Optimization

Traditional SEO optimized for blue links. AI Overview optimization focused on getting cited in summaries.

Answer Engine Optimization (AEO) optimizes for being cited across multiple AI systems—ChatGPT, Perplexity, Claude, and yes, Google’s AI features.

When Google’s own AI systems (AI Overviews vs AI Mode) only share 13.7% citation overlap, the lesson is clear:

Single-platform optimization doesn’t work anymore.

AEO strategies assume citation diversity across systems. They prepare for constant change in how AI platforms select sources.

That approach becomes essential when Google itself can’t decide between AI Overviews and AI Mode.

The Publisher Exodus Nobody Talks About

While Google reduces AI Overviews, publishers are making their own decisions.

Some block AI crawlers entirely (Google-Extended, GPTBot, etc.). Others pursue licensing deals with AI companies. A few embrace AI citations as brand exposure.

But the data suggests a darker trend: publishers optimizing for AI visibility often see traffic declines anyway.

Being cited doesn’t guarantee clicks. And with AI Mode’s 93% zero-click rate, citations might become visibility without value.

The question isn’t whether to optimize for AI features. It’s whether the effort produces meaningful business results when users rarely leave the AI interface.

What Happens in 2026

Three predictions based on the current trajectory:

1. AI Mode Becomes Default in H1 2026

Logan Kilpatrick said “soon” in September 2025. Google has spent Q4 expanding access and building infrastructure. The switch happens in the first half of 2026.

2. AI Overviews Stabilize Around 10-12%

The 15.69% November number will continue declining to roughly 10-12% of queries—primarily informational searches where ad revenue isn’t at stake.

3. Publishers Face Another Traffic Cliff

When AI Mode becomes default, publishers optimized for AI Overviews will discover their citations don’t transfer. Traffic drops again as 93% zero-click behavior becomes the norm.

The Bottom Line Nobody Wants to Admit

Google rolled out AI Overviews to nearly a quarter of all searches, then quietly pulled them back by 36% without explanation.

Meanwhile, they built a completely different system—AI Mode—that cites different sources only 13.7% of the time.

Now that system is being positioned as the “default” experience “soon.”

The rollback wasn’t a failure. It was strategic repositioning.

Google learned that AI Overviews threaten ad revenue on commercial queries. So they’re reducing AI Overviews and replacing them with AI Mode, which keeps 93% of users inside Google.

Publishers who spent 2025 optimizing for AI Overview citations are about to discover that work doesn’t transfer to AI Mode.

And Google never announced any of this. They just quietly changed the rules while everyone else scrambled to adapt.

The future of search isn’t AI Overviews. It’s AI Mode.

And it’s coming soon—whether publishers are ready or not.

Have you seen AI Overview visibility changes in your Search Console data? Reply with your before/after traffic numbers and we’ll analyze the impact.

Did someone forward this to you? Subscribe here for weekly search intelligence Google doesn’t announce.

Ready to optimize for the systems Google won’t explain? SEOengine.ai uses Answer Engine Optimization to get your content cited across AI Overviews, AI Mode, ChatGPT, and Perplexity. Only $5 per article. No monthly commitment. Start here: seoengine.ai

Until next week,

The SEOengine Team

P.S. - When Google’s own AI systems only share 13.7% citation overlap, the message is clear: optimize for multiple AI platforms, not just one. AEO strategies assume constant change. That’s the only approach that survives when Google can’t decide between AI Overviews and AI Mode.

TL;DR: Google quietly updated its core update documentation on December 9, 2025, admitting they “continually make updates to our search algorithms, including smaller core updates” that are “not announced because they aren’t widely noticeable.” This contradicts years of official advice telling SEOs to wait for the next major core update to see ranking recovery. The admission explains “phantom” volatility SEOs tracked throughout 2025—December 7-8, December 3-4, pre-Thanksgiving, November 20, November 12, November 8, Halloween, October 15-17, October 7-8—and why Google only announced 3 core updates in 2025 despite promising 6-8 at the beginning of 2024. The documentation change came just 2 days before the December 2025 core update announcement, buried without any blog post or major statement. SEOs have been gaslighted for years about ranking fluctuations Google claimed didn’t exist.

We caught them.

On December 9, 2025, Google updated their core update documentation.

Not with a blog post. Not with a tweet. Just a quiet documentation change that SEOs almost missed.

Here’s the exact language they added:

“However, you don’t necessarily have to wait for a major core update to see the effect of your improvements. We’re continually making updates to our search algorithms, including smaller core updates. These updates are not announced because they aren’t widely noticeable, but they are another way that your content can see a rise in position (if you’ve made improvements).”

Read that again.

“Including smaller core updates. These updates are not announced.”

Two days later, Google announced the December 2025 core update.

The timing wasn’t coincidental. This was damage control before the SEO community connected the dots.

The Lie They’ve Been Telling For Years

For the past five years, Google’s official guidance has been consistent:

“If your site was negatively impacted by a core update, you may see some recovery between updates. But the biggest change would typically occur after another core update.”

That advice appeared in every official core update blog post. Danny Sullivan repeated it. John Mueller confirmed it. Google’s Search Central documentation stated it clearly.

The message: Wait for the next major core update to see meaningful recovery.

Hundreds of case studies followed this guidance. SEO agencies built recovery strategies around it. Publishers held off on major content overhauls, waiting for the “right” timing.

Meanwhile, their rankings kept fluctuating between major updates.

And Google kept saying those fluctuations were normal volatility, not algorithmic changes.

The Phantom Updates Nobody Could Prove

Barry Schwartz from Search Engine Roundtable has been tracking “unconfirmed Google ranking volatility” for over a decade.

Before Google started announcing core updates in 2016, Barry called them “Phantom updates”—algorithm changes Google never confirmed but that clearly happened.

Even after Google started announcing major updates, the phantom volatility continued.

Here’s what Barry documented in just the final months of 2025:

• December 7-8: Unconfirmed volatility

• December 3-4: Unconfirmed volatility

• Pre-Thanksgiving: “A very big one”

• November 20: Significant movement

• November 12: Ranking shifts

• November 8: The “Movember update”

• Halloween period: Clear volatility

• October 15-17: Unconfirmed changes

• October 7-8: Ranking fluctuations

Every single one of these showed up in third-party tracking tools like SEMrush Sensor, SISTRIX Update Radar, and Mozcast.

Every single one generated chatter in SEO communities.

And every single one, Google never acknowledged.

Until December 9, 2025.

The Promise They Broke

At the beginning of 2024, Google promised the SEO community something specific:

More core updates. More often.

The expectation: 6-8 major core updates per year instead of the previous 3-4.

Here’s what actually happened in 2025:

• March 2025 Core Update (March 13-27)

• June 2025 Core Update (June 30-July 17)

• December 2025 Core Update (December 11-29)

Three updates. Not six. Not eight.

Google broke their promise.

Except now we know they didn’t break it. They just stopped announcing half of them.

The documentation update confirms Google was running “smaller core updates” throughout 2025. Every phantom update Barry tracked was real. Every ranking shift SEOs reported was algorithmic, not “normal volatility.”

Google just chose not to tell anyone.

Why The Timing Matters

December 9, 2025: Documentation updated with the admission.

December 11, 2025: December core update announced.

Two days between admission and announcement.

This wasn’t random timing. This was strategic.

Google knew the December update would generate volatility reports. They knew SEOs would compare it to the “unconfirmed” changes from earlier in December (December 3-4, December 7-8).

By updating the documentation first, they could point to it when questioned: “We already told you we run smaller updates.”

Except they didn’t tell us. They buried it in documentation and hoped nobody would notice the contradiction with five years of previous guidance.

The Gaslighting Timeline

Let’s be clear about what happened:

2020-2024: Google tells SEOs to wait for major core updates to see recovery. Smaller ranking changes between updates are described as “normal volatility” or the result of other factors (spam updates, seasonal changes, etc.).

Throughout 2025: SEOs report consistent ranking volatility between the three major core updates. Google doesn’t acknowledge any of it.

December 9, 2025: Google quietly updates documentation admitting they’ve been running unannounced “smaller core updates” all along.

December 11, 2025: Google announces major core update without addressing the contradiction.

This is textbook gaslighting.

SEOs: “We’re seeing ranking changes.”

Google: “That’s just normal volatility.”

SEOs: “But the tracking tools show—”

Google: “Wait for the next major core update.”

SEOs: “Rankings are changing again.”

Google: “Oh by the way, we’ve been running smaller updates this whole time. We just never mentioned it.”

What The Documentation Actually Says Now

Here’s the full updated section from Google’s core update documentation:

“We understand that a site dropping in search results after a core update is a frustrating experience. But it’s important to understand that the changes we make with these updates aren’t against any specific pages or sites. Instead, the changes are about improving how our systems assess content overall. These changes may cause some pages that were previously under-rewarded to do better.

“One way to think of how a core update operates is to imagine you made a list of the top 100 movies in 2021. A few years later in 2025, you refresh the list. It’s going to naturally change. Some new and wonderful movies that never existed before will now be candidates for inclusion. You might also reassess some films and realize they deserved a higher place on the list than they had before.

“However, you don’t necessarily have to wait for a major core update to see the effect of your improvements. We’re continually making updates to our search algorithms, including smaller core updates. These updates are not announced because they aren’t widely noticeable, but they are another way that your content can see a rise in position (if you’ve made improvements).“

The bold section is new as of December 9, 2025.

It fundamentally changes Google’s stance on ranking recovery timing and algorithm update frequency.

Why This Changes Everything

The admission has three massive implications:

1. Recovery Doesn’t Require Waiting

SEOs have been told for years to wait for the next major core update to see recovery after being hit.

Now Google admits you don’t need to wait. Improvements can show up anytime because they’re constantly running smaller updates.

2. “Normal Volatility” Was Never Just Volatility

Every time SEOs reported ranking changes between major updates, Google implied it was noise—seasonal changes, spam updates, or random fluctuations.

Now we know those changes were often algorithmic. Google was running real core updates and choosing not to announce them.

3. The Tracking Tools Were Right All Along

SEMrush, SISTRIX, Mozcast, and other tracking tools showed clear volatility throughout 2025 between major updates.

Google’s silence implied the tools were picking up false signals.

Now we know the tools were accurate. The signals were real. Google just wasn’t being honest about what was causing them.

The December Update’s Perfect Storm

The December 2025 core update launched December 11 and completed December 29.

It showed two major volatility spikes: December 13 and December 20 (both Saturdays).

But here’s what makes the timing suspicious:

Barry Schwartz reported major unconfirmed volatility on December 7-8 and December 3-4—before Google announced the update on December 11.

Was that volatility from a “smaller core update” running before the major one?

We’ll never know. Because Google doesn’t announce those.

What Publishers Are Saying

The reaction from the SEO community has been swift and angry.

One publisher on Search Engine Roundtable:

“I lost over 90% of my traffic during this core update. I was wiped from Discover, wiped from News, and wiped from Search. 1,000 active users to 10, 5 of them bots from China now. Traffic has been systematically decimated throughout all of 2025. It is 10 steps backwards, 1 step forward and so on.”

Another:

“Traffic systematically decimated throughout 2025” matches exactly what Google just admitted—continuous smaller updates nobody announced.

The pattern is clear: Publishers experienced ongoing algorithmic punishment throughout 2025, not just during the three major updates.

And when they asked Google what was happening, they were told to wait for the next core update.

John Mueller’s Contradictory Statements

On December 13, 2025, John Mueller tweeted:

“We try to launch quality improvements as soon as they’re ready & evaluated. Having quality changes ready for a specific date/time is never a given, and pre-announcing for a fixed date isn’t possible.”

This statement directly contradicts the new documentation.

If Google is “continually making updates” including “smaller core updates,” then they clearly can and do schedule algorithm changes.

The documentation admits these smaller updates happen regularly. Mueller’s tweet implies Google can’t control timing.

Both can’t be true.

The SEO Industry’s Response

The admission has ended years of speculation around phantom volatility.

As one industry publication noted: “Google’s admission about unannounced core updates ends years of speculation around phantom volatility.”

But the reaction isn’t celebration. It’s anger.

SEO agencies built strategies around waiting for major updates. Publishers held off on content improvements, thinking they wouldn’t matter until the next big rollout. Businesses made budget decisions based on Google’s guidance that recovery requires patience.

All of it based on incomplete information Google chose to withhold.

What You Should Do Now

The documentation change creates three immediate action items:

1. Stop Waiting For Major Updates

If you were hit by a core update and waiting to implement changes until “the right time,” stop waiting.

Google just admitted improvements can show up anytime. The old guidance about waiting for major updates was wrong.

2. Monitor Rankings Weekly, Not Quarterly

If Google is running smaller core updates continuously, ranking changes aren’t limited to 3-4 major windows per year.

Your monitoring needs to shift from quarterly check-ins around major updates to weekly tracking of ranking fluctuations.

3. Document Everything

Track every ranking change with detailed notes about what content changes you made and when.

If rankings improve between major updates, you now have proof that smaller updates are real and affecting your site.

The Bigger Pattern

This isn’t the first time Google has quietly changed its story.

On AI Content: Google said for years they could detect AI-generated content and it would be penalized. Then in 2023, they changed their guidance to say AI content is fine if it’s helpful.

On Backlinks: Google minimized the importance of links for years, then leaked documentation showed links are still a top-3 ranking factor.

On User Signals: Google denied using click data for rankings for over a decade. Then the DOJ antitrust case revealed they absolutely do.

The pattern is consistent: Google’s public statements don’t match their actual systems.

The core update admission is just the latest example.

Why This Matters For Answer Engine Optimization

Traditional SEO optimized for Google’s announced algorithm changes.

Answer Engine Optimization (AEO) optimizes for being cited by AI systems that don’t announce changes at all.

If Google now admits they run continuous unannounced updates, the difference between SEO and AEO shrinks.

Both require constant optimization. Both require monitoring changes that aren’t announced. Both require adapting to systems that don’t fully explain how they work.

The shift from “wait for major updates” to “we’re always updating” makes the case for AEO even stronger.

When the algorithm is constantly changing, being citation-worthy matters more than trying to time ranking improvements around specific dates.

The Documentation Nobody’s Talking About

The December 9 update wasn’t the only documentation change Google made.

They also clarified language around how core updates work:

“Core updates adjust the main systems Google uses to rank pages. When those systems get reweighted, every type of page can move, including blogs, service pages, category pages, and product pages.”

This confirms core updates are about reweighting multiple systems, not single-factor changes.

But here’s what’s missing: any explanation of what triggers a “smaller core update” versus a “major core update.”

Google still won’t tell us:

• How often smaller updates run

• What makes an update “small” versus “major”

• Whether smaller updates affect all systems or just some

• How to know if a ranking change came from a smaller update

The admission created more questions than answers.

The Trust Problem

Google’s relationship with the SEO community has always been complicated.

SEOs want transparency. Google provides controlled messaging.

SEOs want detailed guidance. Google gives vague principles.

SEOs want advance warning of changes. Google says that’s impossible.

But the December 9 documentation update crosses a line from “providing limited information” to “contradicting five years of previous guidance.”

When Google tells you to wait for major updates while secretly running smaller ones, that’s not protecting proprietary information.

That’s misleading the people who depend on your platform for their businesses.

What Happens Next

Three things will happen in 2026:

1. More Phantom Updates Will Be Reported

SEO tracking tools will continue showing volatility between major updates.

The difference now: SEOs know those changes are real algorithmic shifts, not noise.

2. Google Will Be Questioned More Aggressively

Every time Google says “wait for the next core update,” SEOs will reference the December 9 documentation.

Google’s credibility on algorithm guidance has been damaged.

3. The Industry Will Adapt

SEO strategies will shift from timing improvements around major updates to continuous optimization.

The playbook just changed from quarterly to weekly.

The Bottom Line

Google ran unannounced algorithm updates throughout 2025 while telling SEOs their ranking changes were “normal volatility” and advising them to “wait for the next major core update.”

On December 9, 2025—two days before announcing the December core update—they quietly admitted the truth in updated documentation.

The admission explains years of phantom volatility, contradicts their official guidance, and proves SEO tracking tools were right all along.

Publishers who lost traffic throughout 2025 weren’t imagining algorithmic changes. Google was running smaller core updates and chose not to announce them.

The SEO community just got confirmation they’ve been gaslighted for years.

And Google did it with a documentation update most people will never read.

Were you tracking phantom volatility in 2025? Reply with your ranking data and we’ll analyze whether smaller core updates affected your site.

Did someone forward this to you? Subscribe here for weekly SEO intelligence that Google doesn’t want you to know.

Want to stop depending on Google’s honesty? Check out how SEOengine.ai optimizes for Answer Engine citations that don’t require waiting for algorithm updates.

Until next week,

The SEOengine Team

P.S. - The next “major” core update typically happens 3-5 months after the last one. That puts it around March-April 2026. But now we know Google will run smaller updates between then and now. Watch your rankings weekly, not quarterly.

The numbers hit like a truck in mid-November 2025.

Salesforce released a report showing AI platforms will drive $263 billion in global online holiday sales. That’s 21% of all holiday orders worldwide.

Not “might drive.” Not “could potentially influence.”

Will drive.

Three weeks later, Amazon quietly updated its robots.txt file and blocked every single OpenAI crawler from accessing its $570 billion marketplace.

ChatGPT’s 700 million weekly users can no longer see Amazon products, prices, or reviews when using Shopping Research.

Meanwhile, Walmart partnered with OpenAI and now receives 36% of its referral traffic from ChatGPT.

This isn’t a feature launch. This is a war for control over the discovery moment before purchase. And the battleground is your product catalog.

The $263 Billion That Changed Everything

Between 40% and 83% of consumers plan to use AI for shopping this holiday season.

Surveys from Visa, Zeta Global, and research firms confirm the same behavior: shoppers now start product searches inside ChatGPT, Perplexity, and Google Gemini.

AI traffic to U.S. retail sites surged 760% between November 2024 and November 2025.

7% of all shoppers now start product searches with AI chat assistants. For Gen Z, that jumps to 10%.

17% of consumers used an AI assistant or LLM for product searches in the past year.

Salesforce projects AI will influence $263 billion in online sales plus $1.6 trillion in in-store sales (through AI-powered recommendations shoppers use while physically shopping).

The data backing this up:

59% of AI-using shoppers will use these tools to find holiday gifts for friends and family.

Shoppers using AI are 30% more likely to make a purchase.

They spend 14% more time browsing.

They generate 8% more revenue per visit.

These aren’t small behavior changes happening at the edges. This is the shopping funnel getting rebuilt from scratch.

What OpenAI Built (And Why Amazon Panicked)

On November 24, 2025, OpenAI launched Shopping Research powered by a specialized GPT-5 mini model trained specifically for shopping tasks through reinforcement learning.

The feature processes 50 million shopping queries daily.

That’s roughly 2% of ChatGPT’s 2.5 billion daily interactions, but it represents concentrated high-intent commercial activity.

How Shopping Research Actually Works:

ChatGPT asks clarifying questions before showing products. Tell it “I need a gift for my four-year-old niece who loves art” and it runs a conversational quiz about budget, interests, and constraints.

Users swipe right on products they like, left on ones they don’t. The AI updates suggestions in real-time based on feedback. It’s Tinder meets Amazon meets personal shopper.

The system takes 3-5 minutes to research across the internet, review quality sources, and synthesize information. This isn’t quick price-checking. It targets considered purchases where users invest time in research.

The accuracy breakthrough:

The model achieves 64% accuracy in matching products to multi-constraint queries, compared to just 37% for standard ChatGPT Search.

What does “multi-constraint accuracy” mean? Ask for “wireless headphones under $150 with active noise cancellation, at least 20 hours battery life, and available in black.” The system verifies recommended products genuinely meet ALL those criteria.

OpenAI states: “We trained it to read trusted sites, cite reliable sources, and synthesize information across many sources to produce high-quality product research.”

The scale nobody expected:

700 million weekly active ChatGPT users globally.

50 million shopping queries processed daily.

“Nearly unlimited usage” offered through the holiday season.

Available to all ChatGPT plans: Free, Plus, and Pro.

This isn’t a pilot program. OpenAI turned on e-commerce at massive scale and retailers had to decide which side they’re on.

Then Amazon Hit the Kill Switch

In late November 2025, e-commerce analyst Juozas Kaziukėnas noticed something in Amazon’s robots.txt file.

New disallow rules for OpenAI crawlers appeared without announcement.

Amazon blocked three critical agents:

1. GPTBot - Scrapes content for ChatGPT model training

2. ChatGPT-User - Fetches live information when users ask questions

3. OAI-SearchBot - Powers OpenAI’s SearchGPT search engine

The result: ChatGPT Shopping Research can no longer access Amazon product pages, prices, specifications, or reviews.

When Modern Retail tested it, asking ChatGPT for specific Amazon product recommendations returned responses like: “Here are some options from other retailers. You can check if they’re available on Amazon yourself.”

Amazon made 40% of U.S. e-commerce invisible to ChatGPT.

Why Amazon Did It (The Brutal Financial Logic):

Amazon generates approximately $56 billion annually from advertising. That business model depends entirely on people browsing Amazon.com.

When shoppers discover products through ChatGPT and buy through OpenAI’s Instant Checkout, Amazon loses:

The discovery moment (worth billions in ad revenue).

The shopping session (where cross-selling happens).

The customer relationship data (what they browsed, considered, abandoned).

The opportunity to promote Amazon’s own brands.

OpenAI collects “a small fee” from each Instant Checkout transaction. That’s revenue Amazon would normally capture.

Scot Wingo, founder of ReFiBuy, captured it perfectly: “Amazon doesn’t want to just be the back end of the internet. They want to be the front door.”

Amazon’s Counter-Strategy: If You Can’t Beat Them, Build Your Own

While blocking external AI agents, Amazon invested heavily in internal tools.

Rufus - Amazon’s AI Shopping Assistant:

Launched in 2024, now with 250 million users.

Just gained the ability to auto-purchase items when prices hit user-defined thresholds.

Amazon says shoppers using Rufus are 60% more likely to complete purchases.

Expected to generate $10+ billion in yearly sales.

Buy For Me:

An agent living inside Amazon’s app that can purchase from OTHER retailers.

If AI shopping is inevitable, Amazon wants to host it on their turf.

Custom Shopping Guides:

Amazon recently tested a feature letting Rufus generate personalized buying guides, directly copying OpenAI’s Shopping Research launched a month earlier.

Amazon CEO Andy Jassy said on the Q3 2025 earnings call: “We’re having conversations with third-party shopping agents and expect we’ll find ways to partner over time.”

Translation: Amazon is negotiating from a position of strength. They’ll participate in AI shopping on their terms, when the economics work.

Until then, they’re building a moat.

Who’s Winning the War (And The Data Proves It)

While Amazon plays defense, competitors went on offense.

Walmart: The Strategic Winner

Announced partnership with OpenAI in October 2025 for direct ChatGPT shopping.

Now receives 36% of referral traffic from ChatGPT, up from 20% initially (Similarweb data).

CEO Doug McMillon: “Agentic AI will be a core growth driver for our e-commerce business.”

One Fortune analyst noted: “Walmart invested years in clean, real-time, end-to-end data. That’s why AI agents can surface Walmart’s product information with confidence. For most retailers, the same query would expose outdated prices, phantom stock, or missed delivery windows.”

Target’s Beta Gambit

Launched app integration inside ChatGPT in November 2025.

Enables purchasing multiple items in single transactions, including groceries.

Offers delivery or curbside pickup selection.

Early data shows 11% increase in interaction rate, 12% improvement in display-to-conversion.

Now gets 15% of referral traffic from ChatGPT.

Etsy’s First-Mover Advantage

Instant Checkout partner since September 2025.

Gets direct access to ChatGPT’s 50 million daily shopping queries.

Receives 20%+ of referral traffic from ChatGPT.

Particularly strong for gift searches where unique, handmade items perform well.

Shopify’s Platform Play

Integrated thousands of merchants including Glossier, Skims, Vuori, and Spanx.

Merchants get ChatGPT visibility without individual partnership deals.

Creates instant distribution channel for small-to-medium brands.

The Numbers Retailers Are Actually Seeing:

Shoppers using AI spend 14% more time browsing.

They’re 30% more likely to make a purchase.

They generate 8% more revenue per visit.

Agencies working with brands report 7x increase in ChatGPT visits after optimization.

Walmart gets 36% of referral traffic from ChatGPT (Similarweb, September 2025).

Target gets 15% referral traffic from ChatGPT.

Etsy gets 20%+ referral traffic from ChatGPT.

eBay gets 10% referral traffic from ChatGPT.

Amazon gets less than 3%, down 18% month-over-month in August 2025.

These aren’t projections. These are measurements of traffic shifts happening right now.

The Uncomfortable Truth About What’s Actually Happening

Here’s what keeps retail executives awake:

McKinsey projects agentic commerce could generate $1 trillion in U.S. retail revenue by 2030.

Morgan Stanley expects 50% of Americans will use AI shopping agents by then.

But the infrastructure is barely functional yet.

Real-world testing reveals constant glitches:

Scot Wingo tried buying a sweater through Perplexity’s Instant Buy. Got error messages despite the product being in stock. Eventually gave up.

Another tester searched for a coffee machine on ChatGPT. It suggested a Breville espresso maker. Clicking through showed an image of a garden rake.

OpenAI’s Instant Checkout only works for single-item purchases.

You can’t connect loyalty programs like Walmart+ yet.

Inventory accuracy remains inconsistent.

OpenAI acknowledges in their documentation that Shopping Research “may make mistakes about product details like price and availability” and encourages users to verify on merchant sites.

The 64% accuracy rate on multi-constraint queries means nearly half of complex recommendations contain errors or outdated information.

This is Version 1.0 of AI shopping. It’s clunky, limited, and frustratingly inconsistent.

But 50 million people are using it every day anyway.

Why Reddit Reviews Now Outrank Your Product Pages

OpenAI representatives briefing reporters stated explicitly: “User experiences shared on Reddit may be considered more trustworthy than paid marketing or reviews posted on a product page.”

This isn’t a bug. It’s the core feature.

Isa Fulford, who leads OpenAI’s Shopping Research team, acknowledged during the press briefing that teaching the model to identify objective, unpaid reviews has been “a pretty hard task” and that it’s “impossible to get it 100% correct.”

The solution? Default to sources less likely to contain paid content.

As OpenAI researcher Manuka Stratta stated directly: “Generally a lot of reviews on Reddit are pretty trustworthy.”

This creates an unprecedented dynamic:

A single Reddit thread discussing your product now carries more weight in ChatGPT’s recommendations than your meticulously optimized product pages.

When Michael Wieder, co-founder of baby retailer Lalo, restructured his product listings, he moved beyond keyword stuffing with basic attributes like material, size, and color.

Instead, he added detailed information like “good for small spaces,” “great for apartment living,” or “best gifts for kids under one year old.”

“I’m looking for the best gift for a child that is, you know, this age that lives in this place,” Wieder told CNBC. “We’re taking it a step further in how we construct the infrastructure of our website and the content that lives within our website.”

PacSun publicly stated it hopes to join OpenAI’s platform and is reformatting its website for AI search visibility.

Brands are redirecting budgets from SEO to AEO (Answer Engine Optimization).

One agency executive said: “We’ve seen a major surge in demand from retailers and brands that have started to see a steep decline in traffic from social media ads and search engines.”

The Three Paths Forward (And Which One Kills Your Business)

If you sell products online, the AI shopping revolution creates three distinct scenarios.

Path 1: Amazon-Exclusive (You Just Became Invisible)

You’re invisible to 700 million ChatGPT users and 50 million daily shopping queries.

Your options:

Diversify to Shopify/DTC to capture ChatGPT visibility.

List on Etsy for immediate Instant Checkout access.

Launch Walmart marketplace presence.

Accept the trade-off: Amazon sales vs. AI discovery.

Example: Brands exclusively on Amazon forfeit AI visibility entirely until Amazon unlocks access to ChatGPT (which may or may not happen).

Path 2: Multi-Channel (You’re Positioned to Win)

Optimize for:

Structured product data (JSON-LD, schema markup).

GTIN/MPN identifiers on all products.

Benefit-focused product descriptions answering “who is this for?”

Community presence (Reddit discussions, review platforms).

Crawler access verification (ensure OAI-SearchBot can access your site).

Example: Walmart’s clean, real-time data infrastructure allows AI agents to surface their products confidently. Target’s multi-constraint product metadata enables accurate recommendations.

Path 3: DTC/Shopify (You Have the Biggest Opportunity)

ChatGPT Shopping prioritizes “trusted sites” and community discussion over commercial content.

Focus on:

Getting featured in Reddit discussions about your product category.

Building FAQ schema for common questions.

Creating use-case content (”How to choose X for Y”).

Answering multi-constraint queries explicitly in product descriptions.

Example: Brands getting featured in Reddit threads about their product category see disproportionate recommendations in ChatGPT Shopping Research.

What Happens When Google Joins the Fight

Google isn’t sitting this out.

Google Gemini, Google AI Overviews, and Google Shopping all compete for the same AI-driven discovery moment.

Google has direct access to product feeds submitted via Google Merchant Center Next, allowing for extensive real-time product information.

ChatGPT Shopping has a far more limited dataset and doesn’t provide a browsing experience beyond the initial product selection.

When Brodie Clark tested ChatGPT Shopping versus Google Shopping, he concluded: “ChatGPT Shopping still has a long way to go before it is truly useful to shoppers.”

Google AI Overviews reach 2 billion monthly users across 200+ countries, according to CEO Sundar Pichai. Gemini App has 450 million monthly active users.

The battleground isn’t just ChatGPT versus Amazon. It’s every AI platform competing for the discovery moment before purchase.

Why SEOengine.ai Matters More Now Than Ever

Traditional SEO optimized for Google search rankings.

AEO (Answer Engine Optimization) optimizes for being cited by AI systems.

The difference? Google ranks pages. AI systems cite sources.

SEOengine.ai uses a multi-agent AI system with five specialized agents that perform competitor analysis, mine human context from Reddit/YouTube/LinkedIn/X.com, conduct research verification, replicate brand voice, and optimize for both traditional SEO and Answer Engine Optimization.

The platform generates 4,000-6,000 word articles optimized for ranking in ChatGPT, Perplexity, Google AI Overviews, and traditional search engines.

Key differentiators:

90% brand voice accuracy versus competitors’ 60-70%.

8/10 content quality in bulk mode compared to industry average 4-6/10.

First-mover advantage in AEO optimization for AI search engines.

Publication-ready content requiring minimal editing.

Pricing that makes sense:

Pay-As-You-Go: $5 per post (after discount) with no monthly commitment.

Unlimited words per article.

Bulk generation available (up to 100 articles simultaneously).

All features included (AEO optimization, brand voice, SERP analysis, WordPress integration).

No hidden fees or credit systems.

When 40% of U.S. e-commerce is invisible to ChatGPT and brands are redirecting budgets from SEO to AEO, having content optimized for AI citation becomes the difference between being recommended or being invisible.

The Real Competition Isn’t Amazon vs. Walmart. It’s Perplexity.

While retailers fight over ChatGPT visibility, Perplexity launched its own AI shopping assistant with PayPal Instant Buy integration.

The feature launched in late November 2025, positioning Perplexity as the “answer engine” for shopping.

Perplexity’s differentiators:

Contextual memory across shopping sessions. Ask for a jacket for ferry commutes to San Francisco, and it remembers that context when you later search for boots.

Merchant-friendly approach. Retailers remain merchant of record. Perplexity doesn’t cut them out of the customer relationship.

Real-time web search and source attribution. While ChatGPT relies on training data that can become outdated, Perplexity pulls live information from across the web.

PayPal integration for instant checkout directly in chat.

The catch:

When Fortune tested Instant Buy, it wasn’t available on Enterprise accounts. On free personal accounts, only a handful of brands offered the Instant Buy option.

A Perplexity spokesperson said the feature would roll out to many more products and brands over the next few weeks.

Early testing showed error messages even for in-stock products. One tester tried purchasing from Abercrombie & Fitch and eventually gave up after repeated failures.

Why it matters:

Between 15% and 30% of online shoppers are expected to use generative AI to shop for holiday gifts, according to a Bain survey.

The race isn’t about which platform has the best AI. It’s about which platform captures the discovery moment before purchase and converts that into completed transactions.

Google dominates search. Amazon dominates e-commerce. ChatGPT has 700 million users. Perplexity has real-time data and merchant partnerships.

The winner isn’t decided by technology. It’s decided by who solves the highest friction point in the buying journey.

The Three Questions Every Brand Must Answer Right Now

Question 1: Where does your brand show up in AI shopping results?

Test it yourself. Open ChatGPT, Perplexity, and Google Gemini.

Search for products in your category. See which brands appear. See if yours does.

If you’re not showing up, you’re losing discovery traffic right now.

Question 2: Can AI agents access your product data?

Check your robots.txt file. Verify you’re not accidentally blocking AI crawlers.

Ensure OAI-SearchBot, ChatGPT-User, and CCBot (Common Crawl) can access your site.

Add proper schema markup (Product, FAQPage, HowTo).

Question 3: Are you Amazon-exclusive?

If yes, you just became invisible to 700 million ChatGPT users.

Amazon will eventually partner with AI platforms. But “eventually” could be 6 months. It could be 2 years.

How much discovery traffic are you willing to lose while you wait?

Why Amazon’s Strategy Might Backfire (And What Happens Next)

Amazon faces what retail consultants call “the leader’s dilemma.”

Their market share is so significant (40% of U.S. e-commerce) that they have the most to lose from AI shopping agents.

But blocking AI agents creates its own risks:

Brands are diversifying:

PacSun publicly stated it hopes to join OpenAI’s platform and is reformatting its website for AI search visibility.

Other brands are redirecting budgets from SEO to AEO.

One agency executive: “We’ve seen a major surge in demand from retailers and brands that have started to see a steep decline in traffic from social media ads and search engines.”

The discovery gap:

When 7% of shoppers now start product searches with AI tools (10% for Gen Z), being invisible means losing the consideration moment entirely.

Competitive arbitrage:

Walmart, Target, and thousands of Shopify merchants are capturing traffic Amazon isn’t even competing for.

That gap grows daily.

The retention problem:

Amazon’s subsidiaries (Zappos, Shopbop, Woot) don’t block AI agents in their robots.txt files. Amazon is experimenting through smaller properties before risking its crown jewel.

Jassy’s comment about “finding ways to partner” suggests the company knows this isn’t sustainable long-term.

They’re buying time to build internal solutions while the external AI shopping infrastructure matures.

But time might be the one thing Amazon doesn’t have.

What To Watch In 2026

Three developments will determine how this war plays out:

Development 1: ChatGPT Sponsored Placements

OpenAI currently states results are “organic and unsponsored, ranked by relevance.”

But processing 50 million free shopping queries daily isn’t sustainable without monetization.

Sponsored product placements in ChatGPT Shopping Research are inevitable. The question is timing and how OpenAI structures the ad model.

Development 2: Amazon’s Partnership Decision

Jassy said Amazon is “having conversations” with AI shopping agents.

When (not if) Amazon unlocks access to ChatGPT, the terms of that partnership will reshape retail economics.

Will Amazon pay for placement? Take a revenue share? Require exclusive partnerships?

The deal structure sets precedent for the entire industry.

Development 3: Accuracy Improvements

The 64% accuracy rate will improve. When it hits 85-90%, consumer trust will accelerate adoption dramatically.

Track OpenAI’s model updates. Each improvement shifts more shopping from traditional search to AI discovery.

The $263 billion Salesforce projects for 2025 holiday shopping is just the beginning.

That number grows every quarter as accuracy improves, friction decreases, and consumer behavior shifts.

Frequently Asked Questions

What is ChatGPT Shopping Research and how does it work?

ChatGPT Shopping Research is a specialized AI-powered shopping assistant launched by OpenAI on November 24, 2025. It processes 50 million daily shopping queries from 700 million weekly ChatGPT users. The system uses a reinforcement-trained variant of GPT-5 mini that asks clarifying questions, researches across the internet, and generates personalized buyer’s guides in 3-5 minutes. Users swipe on product recommendations to refine results in real-time.

Why did Amazon block ChatGPT from accessing its website?

Amazon blocked all OpenAI crawlers (ChatGPT-User, OAI-SearchBot, GPTBot) in late November 2025 to protect its $56 billion annual advertising business. When shoppers discover products through ChatGPT, Amazon loses the discovery moment, cross-selling opportunities, customer relationship data, and the ability to promote its own brands. OpenAI also collects fees from Instant Checkout transactions, capturing revenue Amazon would normally earn.

How much revenue does AI shopping actually generate?

Salesforce projects AI will drive $263 billion in global online holiday sales in 2025, representing 21% of all holiday orders. This includes both direct purchases and AI-influenced buying decisions. Additionally, AI drives $1.6 trillion in in-store sales through recommendations shoppers use while physically shopping. McKinsey projects agentic commerce could generate $1 trillion in U.S. retail revenue by 2030.

What percentage of shoppers actually use AI for shopping?

Between 40% and 83% of consumers plan to use AI for shopping this holiday season, according to surveys from Visa and Zeta Global. 7% of all shoppers now start product searches with AI chat assistants, jumping to 10% for Gen Z consumers. 17% of consumers used an AI assistant or LLM for product searches in the past year. Nearly 60% of U.S. consumers have used a generative AI tool for help with online shopping.

How much traffic does ChatGPT send to retailers?

Walmart receives 36% of its referral traffic from ChatGPT (up from 20% initially). Target gets 15% of referral traffic from ChatGPT. Etsy receives 20%+ of referral traffic from ChatGPT. eBay gets 10% of referral traffic from ChatGPT. Amazon receives less than 3% of referral traffic from ChatGPT, down 18% month-over-month in August 2025. Agencies report brands see a 7x increase in ChatGPT visits after optimization.

Can you actually buy products directly in ChatGPT?

OpenAI’s Instant Checkout feature allows purchases directly within ChatGPT for select partners including Walmart, Target, Etsy, and thousands of Shopify merchants (Glossier, Skims, Vuori, Spanx). The feature launched in September 2025 and is rolling out gradually. Currently, users can only purchase one item at a time and cannot connect loyalty memberships like Walmart+. Instant Checkout is not yet connected to Shopping Research.

How accurate is ChatGPT Shopping Research?

ChatGPT Shopping Research achieves 64% accuracy in matching products to multi-constraint queries, compared to 37% for standard ChatGPT Search. OpenAI acknowledges the system “may make mistakes about product details like price and availability” and encourages users to verify information on merchant websites. The system works best for products with lots of specifications like camping gear, backpacks, or musical equipment.

What’s the difference between ChatGPT and Perplexity for shopping?

ChatGPT Shopping Research focuses on deep product research, taking 3-5 minutes to generate comprehensive buyer’s guides. Perplexity focuses on quick discovery with contextual memory across sessions and offers PayPal Instant Buy for immediate purchases. While ChatGPT has 700 million weekly users, Perplexity pulls real-time web data versus relying on training data. Perplexity keeps merchants as merchant of record, preserving customer relationships.

Why do Reddit reviews matter more than product page reviews?

OpenAI explicitly stated that “user experiences shared on Reddit may be considered more trustworthy than paid marketing or reviews posted on a product page.” The Shopping Research model defaults to sources less likely to contain paid content. OpenAI researcher Manuka Stratta confirmed: “Generally a lot of reviews on Reddit are pretty trustworthy.” A single Reddit thread discussing your product now carries more weight than optimized product pages.

What should brands do if they’re Amazon-exclusive?

Brands exclusively on Amazon are currently invisible to ChatGPT’s 50 million daily shopping queries. Options include: diversifying to Shopify/DTC for ChatGPT visibility, listing on Etsy for Instant Checkout access, launching Walmart marketplace presence, or accepting the trade-off between Amazon sales and AI discovery while waiting for Amazon to eventually partner with AI platforms.

How does Amazon’s Rufus compare to ChatGPT Shopping?

Amazon’s Rufus has 250 million users and integrates directly with Amazon’s catalog. Shoppers using Rufus are 60% more likely to complete purchases. Rufus recently gained the ability to auto-purchase items when prices hit user-defined thresholds and can generate custom shopping guides. Amazon also tests “Buy For Me,” an agent that can purchase from other retailers directly in Amazon’s app.

What is Answer Engine Optimization (AEO)?

Answer Engine Optimization (AEO) optimizes content for being cited by AI systems like ChatGPT, Perplexity, and Google Gemini. Unlike traditional SEO which focuses on ranking pages, AEO focuses on getting cited as a trusted source. Key elements include structured data, FAQ schema, benefit-focused product descriptions, community presence on Reddit, and ensuring AI crawlers can access your site.

How much does ChatGPT Shopping cost to use?

ChatGPT Shopping Research is available for free to all ChatGPT users including Free, Plus, and Pro plans. OpenAI offered “nearly unlimited usage” through the 2025 holiday season. Instant Checkout purchases follow standard merchant pricing with no additional ChatGPT fees to consumers. OpenAI collects “a small fee” from merchants for each transaction completed through Instant Checkout.

Will Amazon eventually partner with ChatGPT?

Amazon CEO Andy Jassy stated on the Q3 2025 earnings call: “We’re having conversations with third-party shopping agents and expect we’ll find ways to partner over time.” Amazon is currently negotiating from a position of strength, building internal AI tools like Rufus and Buy For Me while evaluating partnership economics. The timing remains uncertain but industry analysts expect eventual integration.

How do I check if AI can access my product pages?

Check your robots.txt file at yoursite.com/robots.txt. Verify you’re not blocking OAI-SearchBot (OpenAI), ChatGPT-User, GPTBot, CCBot (Common Crawl), or PerplexityBot. Add these user agents to your robots.txt with “Allow: /” to grant access. Verify proper schema markup using Google’s Rich Results Test. Test by searching for your products in ChatGPT Shopping Research.

What conversion rates do AI shopping tools actually achieve?

Shoppers using AI are 30% more likely to make a purchase, spend 14% more time browsing, and generate 8% more revenue per visit. Target reported an 11% increase in interaction rate and 12% improvement in display-to-conversion after launching ChatGPT integration. Early research shows LLM referrals currently convert worse than Google traffic but improve month-over-month as the technology matures.

Is ChatGPT Shopping better than Google Shopping?

Google Shopping has direct access to product feeds via Google Merchant Center Next, providing extensive real-time product information. ChatGPT Shopping has a more limited dataset and doesn’t provide browsing beyond initial recommendations. Industry testing concludes “ChatGPT Shopping still has a long way to go before it is truly useful.” Google AI Overviews reach 2 billion monthly users versus ChatGPT’s 700 million weekly users.

What’s the biggest mistake brands make with AI shopping?

The biggest mistake is staying Amazon-exclusive while Amazon blocks AI crawlers. Brands forfeit visibility to 700 million ChatGPT users and 50 million daily shopping queries. Other critical mistakes include blocking AI crawlers accidentally in robots.txt, lacking proper schema markup, focusing only on keyword-stuffed product descriptions rather than benefit-focused content, and ignoring community presence on Reddit and forums.

How does SEOengine.ai help with AI shopping visibility?

SEOengine.ai uses a multi-agent AI system optimizing content for both traditional SEO and Answer Engine Optimization (AEO). The platform generates 4,000-6,000 word articles optimized for ChatGPT, Perplexity, Google AI Overviews, and search engines. Key features include 90% brand voice accuracy, competitor analysis, Reddit/YouTube/LinkedIn context mining, and pricing at $5 per post with no monthly commitment.

What happens to Amazon sellers when ChatGPT scales?

Amazon sellers face a critical decision: maintain Amazon exclusivity and forfeit ChatGPT visibility, or diversify distribution to capture AI-driven discovery. Some brands will prioritize Amazon’s massive existing customer base. Others will pursue multi-channel strategies capturing both Amazon sales and AI discovery traffic. The outcome depends on how quickly Amazon partners with AI platforms and what terms they negotiate.

The Bottom Line That Nobody Wants to Admit

AI shopping isn’t coming. It already happened.

50 million queries daily. 700 million weekly users. $263 billion in holiday sales influenced by AI recommendations.

Amazon can block OpenAI’s crawlers, but it can’t block the behavioral shift happening across retail.

When Walmart gets 36% of referral traffic from ChatGPT and Target sees double-digit increases in conversion metrics, the data speaks louder than any press release.

The question isn’t whether AI shopping will disrupt e-commerce.

The question is whether your business will be visible when it does.

Because right now, 40% of U.S. e-commerce is invisible to 700 million ChatGPT users.

And that number changes every time a retailer decides which side of history to be on.

The retailers winning this war aren’t the ones with the biggest AI budgets. They’re the ones who recognized that when Google killed keyword stuffing, survival meant adapting to semantic search.

When mobile traffic surpassed desktop, survival meant responsive design.

When AI becomes the discovery layer above search, survival means being citation-worthy rather than ranking-worthy.

SEOengine.ai exists because the rules changed. Traditional SEO optimized for Google’s algorithm. Answer Engine Optimization optimizes for being cited by AI systems as a trusted source.

The brands thriving in 2026 won’t be the ones fighting the AI shopping wave. They’ll be the ones riding it.

Where does your brand show up in AI shopping results today?

Because if the answer is “nowhere,” your competitors just got a head start you can’t afford to give them.

Ready to dominate AI shopping visibility? SEOengine.ai generates AEO-optimized content that gets cited by ChatGPT, Perplexity, and Google Gemini. Pay-as-you-go pricing at $5 per article. No monthly commitments. Try it now at SEOengine.ai.

Want to test your AI visibility? Search for your product category in ChatGPT, Perplexity, and Google Gemini. See which brands appear. Then contact us to fix what’s broken.

Need the data that backs this up? Download our complete competitive analysis showing referral traffic shifts, conversion rates, and merchant strategies across AI shopping platforms.

We’ve been tracking Google updates since 2015, but the December 2025 core update broke every pattern we thought we understood about search.

The update officially finished rolling out on December 29, 2025 at 2:05 PM ET—exactly 18 days and 2 hours after it started. Google announced it was “complete” via their Search Status Dashboard, and the SEO community exhaled.

But here’s what doesn’t make sense:

A documented case study shows a site with impressions UP 27.56% year-over-year, average rankings IMPROVED by 14.01%, but clicks DOWN 36.18% and click-through rate collapsed from 5.98% to 3.35%.

You read that right. Better rankings. Lower traffic.

That’s not how SEO is supposed to work.

The Numbers Nobody Expected

Google announced the December 2025 core update on December 11 at 12:25 PM Eastern. This was the third core update of 2025, following March and June updates.

The industry tracking tools lit up immediately:

• Semrush Sensor hit 8.7/10 volatility

• SISTRIX Update Radar showed elevated movement starting December 7-8 (before the official announcement)

• Two massive volatility spikes: December 13 and December 20 (both Saturdays)

But the real story wasn’t in the tracking tools. It was in what happened to real businesses.

Global impact by the numbers:

• 40-60% of websites affected worldwide

• Average organic traffic decline: 17.3%

• Average ranking drop: 3.2 positions

• 12% of sites: traffic losses exceeding 70%

Industry-specific devastation:

• E-commerce sites: 52% impacted

• Health/YMYL content: 67% affected

• Affiliate sites: 71% hit with losses

• Mass-produced AI content: 87% negative impact

One publisher on Search Engine Roundtable wrote: “I lost over 90% of my traffic during this core update. I was wiped from Discover, wiped from News, and wiped from Search. 1,000 active users to 10, 5 of them bots from China now. Slashed in a matter of hours.”

That’s not an algorithm adjustment. That’s a business extinction event.

The Google Discover Massacre

If there’s one defining characteristic of the December update, it’s what happened to Google Discover traffic.

Google Discover had quietly become the dominant traffic source for news publishers throughout 2025. Research published in August showed Discover accounting for two-thirds of Google referrals for news and media websites.

Then December 12 happened.

Publishers reported:

• 70-85% traffic declines in daily visitor counts

• 98% drops in Discover impressions in the days before the announcement

• Complete elimination of Discover traffic within 24-48 hours

• 750,000 impressions disappearing overnight

Sites that had stable Discover performance for years went to absolute zero in 24 hours. No warnings. No manual actions. No quality issues. Just binary elimination from distribution.

Approximately 9,000 UK-based sites experienced significant Discover disruption according to industry analysis.

The timing? Brutal. December is when seasonal advertising rates peak. Publishers faced their worst traffic period in years during what should have been their most lucrative month.

The Ranking Paradox Explained

So how do you lose traffic when your rankings improve?

The answer lies in how Google now counts impressions versus how users actually click.

Google counts impressions separately for BOTH AI Overviews and traditional organic results. This means if you rank on page one AND appear in the AI Overview for the same query, you receive TWO impressions.

This double-counting inflates impression metrics dramatically. But clicks don’t increase proportionally because users find their answers in the AI Overview without ever clicking through to any website.

The result:

Impressions surge. Rankings improve. Clicks crater.

Even queries WITHOUT AI Overviews experienced 41% organic CTR declines year-over-year, suggesting users are clicking less across the board.

Industry data from Seer Interactive’s September 2025 study reveals the scale of this problem:

• Organic CTR plummeted 61% for queries with AI Overviews (from 1.76% to 0.61%)

• Paid CTR crashed 68% (from 19.7% to 6.34%)

But there’s a survival path: brands cited in AI Overviews earned 35% more organic clicks and 91% more paid clicks than non-cited competitors.

Getting cited became more valuable than ranking #1.

Who Survived (And How)

Not everyone lost. Some sites saw dramatic gains during the December update.

Winners showed specific patterns:

• Sites with deep content clusters (10-15 high-quality supporting articles): +23% visibility

• Authoritative health sites with licensed authors and citations: +42% traffic

• Niche experts with original testing and proprietary data: +38% gains

• Content featuring first-hand research: +47% increases

One retail site, Dunelm, gained 17% during the June 2025 update and nearly doubled their organic visibility in following months.

The pattern is clear: sites built on depth, expertise, and original insight gained ground. Sites built on volume, templates, and generic content got crushed.

Technical performance mattered more than ever:

• Sites with Largest Contentful Paint under 1.6 seconds maintained or improved rankings in 71% of cases

• Poor Interaction to Next Paint scores (>300ms) correlated with 31% more traffic loss

• Mobile CLS issues proved particularly devastating

What This Means For 2026

The December 2025 core update didn’t just shuffle rankings. It redefined what “winning” looks like in search.

Three fundamental shifts happened:

1. Distribution trumps rankings Your position doesn’t matter if Google doesn’t distribute your content. Discover, News, and even traditional search can simply stop serving your pages regardless of quality signals.

2. Citations beat clicks Being referenced in AI Overviews generates more qualified traffic than ranking #3 in traditional results. The game shifted from “rank higher” to “get cited more.”

3. Visibility is volatile Sites that were stable for years disappeared in hours. The buffer that used to exist between “good enough” and “not good enough” has vanished.

Google confirmed in updated documentation that they release smaller, unannounced core updates continuously. This explains the unexplained volatility throughout 2025 that tracking tools missed.

The implication: You can lose traffic between major updates and never know why.

What You Should Do Now

If you were hit by the December update, here’s what recovery data shows:

Immediate actions (Week 1-2):

• Confirm the impact using Google Search Console data comparing December 1-10 vs December 12-31

• Identify which pages lost visibility (don’t react to overall traffic drops)

• Check if you lost Discover traffic specifically vs general search traffic

Short-term strategy (Months 1-3):

• Audit content for depth—thin coverage saw consistent losses

• Add original research, proprietary data, first-hand testing to top pages

• Strengthen E-E-A-T signals: author credentials, citations, transparent sourcing

• Fix Core Web Vitals issues (LCP, INP, CLS all weighted more heavily)

Long-term positioning (Months 3-6):

• Build content clusters around core topics (10-15 interconnected deep articles)

• Stop publishing volume-based content—quality per piece matters infinitely more

• Develop citation-worthy insights that AI systems want to reference

• Diversify beyond Google search (email, social, direct traffic)

Recovery typically takes 2-6 months for most sites. YMYL content requires 6-12 months.

Some sites won’t recover at all. That’s the new reality.

The Uncomfortable Truth

The December 2025 core update exposed what many suspected but few wanted to admit: Google’s AI features are fundamentally incompatible with the click-based ecosystem that’s supported digital publishing since the early 2000s.

When traffic declines approach 90% for publishers who followed every quality guideline, we’ve moved beyond “optimization challenges” into existential crisis territory.

One Search Engine Roundtable commenter captured it perfectly: “Traffic has been systematically decimated throughout all of 2025. This latest core update has been positive but it doesn’t matter. It is 10 steps backwards, 1 step forward and so on.”

The December update finished on December 29. But the disruption it represents has just begun.

If you’re seeing unusual traffic patterns, weird ranking movements, or inexplicable drops in impressions—you’re not alone. And you’re not imagining it.

Google’s search is changing faster than most businesses can adapt.

The question isn’t whether you’ll be affected by the next update.

The question is whether you’ll still be in business when it hits.

What’s your December update story? Reply to this email and let us know if you survived, struggled, or found an unexpected win. We read every response.

Did someone forward this to you? Subscribe here to get SEO intelligence that actually matters, delivered weekly.

Want to dig deeper? Check out our analysis of why AI Overviews are fundamentally breaking the publisher business model.

Until next one,

The SEOengine Team

P.S. - The next major Google update typically happens 3-5 months after a core update. That puts us around March-April 2026. Start preparing now.