The SEO-AEO gap: why Google rankings no longer predict AI citations

Only 12% of URLs cited by ChatGPT, Perplexity, and Copilot rank in Google's top 10 for the original query. Eighty percent don't rank in the top 100. After a decade of measuring SEO success by page-one positions, the metric that used to predict your discoverability now predicts fewer than one in eight AI citations. The gap is real, it's measurable, and it reshapes where marketing teams should be pointing their effort in 2026.

Why this matters

AI Overviews now appear on roughly 25.8% of U.S. queries and 88-91% of informational searches. When they do, Ahrefs and Semrush both measured a 34.5% drop in click-through for the #1 ranked result. Pew's 68,000-query sample found a 46.7% relative decline (8% with AIO vs 15% without). The functional experience of "I ranked first for this query" is quietly getting cut by a third or more, and the slack is going to a different discovery loop: AI answers that draw from their own set of citations, most of which are not your page-one Google result.

The SEO playbook your team built on - rank for the head term, wait for traffic - is still useful for brand defense, navigational intent, and deep-funnel queries. It is no longer the same asset it was for mid-funnel discovery. AI answer engines have become an adjacent channel with their own ranking system, their own weight functions, and their own winners.

Method

This insight is a synthesis of six public studies published Q3 2025 - Q1 2026, triangulated against qualitative observations from Passionfruit's AEO monitoring work across B2B SaaS and D2C e-commerce clients in Q1 2026. Every number in the body is attributed to its source below.

The public studies:

• Ahrefs AI search overlap study (August 2025): 15,000 long-tail queries run through Google, Bing, and AI assistants. 1.9M citations from 1M AI Overviews analyzed.
• Ahrefs AI Overviews CTR study (2025): 300,000 keywords, measuring #1 CTR impact when AI Overviews are present.
• Semrush AI Overviews study (2025-2026): 10M+ keywords, AIO impact on organic click-through.
• Pew Research AI Overviews study (2025): 68,000 queries, AIO click behavior. Widely referenced across industry analyses.
• Conductor AI Overviews prevalence analysis (2026): 21.9M queries. Used as the base rate for AIO appearance.
• Reddit citation share tracking from Ahrefs, Similarweb, and SaaS Intelligence (Oct 2025 - Jan 2026): platform-by-platform Reddit share across ChatGPT, Perplexity, and Gemini.

Limitations of this approach: the six studies use different methodologies, samples, query mixes, and timeframes. The headline 12% overlap number is specific to ChatGPT, Perplexity, and Copilot on long-tail queries sampled in August 2025. AI Overviews behave differently: a January 2026 Ahrefs follow-up found 38% of AIO citations come from the top 10 - closer to SERP alignment, because AIO runs on top of Google's own index. We call out platform specifics in each finding.

Our own observations are qualitative. We did not run a controlled study; we referenced operational patterns seen across client AEO programs, redacted of any brand-specific data. For a deeper look at ChatGPT-specific citation share across agentic SaaS queries, see our companion insight on AEO citation share in ChatGPT.

Six findings

1. Google's top 10 predicts 12% of AI citations

Ahrefs' August 2025 study remains the cleanest dataset on overlap. Across 15,000 long-tail queries asked of ChatGPT, Perplexity, and Copilot, only 12% of cited URLs ranked in Google's top 10 for the original prompt. Eighty percent didn't rank in the top 100. The page that wins in Google is usually not the page that wins in an AI answer. They are pulling from different corners of the index, using different retrieval signals.

Worth separating: AI Overviews, Google's own native AI feature, tracks closer to SERP. A January 2026 follow-up put AIO overlap at 38% - about three times the external-LLM number, but still well under half. If your program measures SEO and AEO as a single funnel, the AIO overlap is the only subset where that framing roughly holds.

2. Perplexity is the outlier

Perplexity's citation set overlaps with Google's top 10 at roughly 30%, about one in three. That's double ChatGPT's overlap and nearly triple Gemini's. Perplexity also cites the most sources per response (average 21.87), versus Claude's 5.67 and Copilot's 6.89. If you treat AEO as a single target, you miss the shape of the problem: Perplexity is closer to an SEO redistribution, while ChatGPT and Gemini are adjacent channels. Per one recent analysis: Gemini trusts what your brand says; ChatGPT trusts what the internet agrees on; Perplexity trusts industry experts and customer reviews.

3. Citation sets are narrower than SERPs

Google's top 10 surfaces around 10 domains per query. AI answers typically pull from 3 to 6. Roughly 30 domains capture 67% of citations within a topic. Wikipedia and Reddit each account for 12-13% of ChatGPT U.S. citations alone. The winners' circle is structurally smaller than it was on a SERP.

ChatGPT narrowed further when its default model switched to GPT-5.3 Instant. The average unique domains per answer dropped from 19.1 to 15.2, a 20% tightening, per industry tracking. Translation: the winners' circle is smaller than it was a year ago, and getting smaller.

4. Reddit is a platform-specific lever

The most visible fragmentation case: Reddit citation share in January 2026 was 5% on ChatGPT, 24% on Perplexity, and 0.1% on Gemini. Across all tracked categories, Reddit's ChatGPT share grew at least 73% from October 2025 to January 2026; in several purchase-adjacent verticals it more than doubled.

This is not "Reddit matters." It is: Reddit matters on ChatGPT and Perplexity, and barely at all on Gemini. A generic "invest in Reddit" AEO strategy over-allocates on Gemini-weighted programs and under-allocates on Perplexity-heavy ones. The playbook has to be platform-specific.

5. AI Overviews are cutting #1 clicks by a third

Across two independent large-sample studies, Ahrefs (300k keywords) and Semrush (10M+ keywords), the CTR for the #1 organic result when AI Overviews appear drops 34.5%. Pew's 68,000-query study found a 46.7% relative decline (8% with AIO vs 15% without). Amsive Digital's 700k-keyword sample measured a more modest 15.49% drop.

The range is real (AIO impact varies by query type and AIO layout), but the direction is unanimous. AIO is present on 25-26% of U.S. queries as of early 2026 and appears on 88-91% of informational searches. For any team whose mid-funnel strategy leans on informational content, the AIO cut is not a tail risk. It is the base rate. We wrote about a related ceiling effect on template-driven content in the pSEO efficiency curve.

6. What the clicks that do leave look like

The studies don't prove it, but Passionfruit's operational monitoring across B2B SaaS and D2C clients in Q1 2026 matches three qualitative patterns reported in the industry analyses.

• Listicles, articles, and product pages dominate citations. Listicles at 21.9%, articles at 16.7%, and product pages at 13.7% across ChatGPT, Perplexity, and AI Mode. Documentation, FAQ pages, and question-shaped landing pages are over-represented relative to their share of Google results.
• Brand-name queries are the cleanest wins. When the user asks about a specific brand, LLMs and Google agree more often. Where Google and the LLMs diverge is the mid-funnel "best X for Y" or "how does X compare to Y" prompts.
• Comment-thread content shows up where Reddit is indexed. On ChatGPT and Perplexity specifically, answer snippets often quote Reddit comments directly, not Reddit posts. The operational implication: your brand needs a comment strategy, not just a post strategy.

What this means for marketers

Stepping back from the numbers:

• Treat AEO as a different channel, not a rebadge of SEO. A top-10 Google ranking is a 12% signal, not a win. Your measurement, target list, and publishing loop all need to be rebuilt around citations, not SERP positions.
• Build platform-specific strategies. Perplexity, ChatGPT, Gemini, and AI Overviews have different trust models and different citation sets. Consolidating them into "AI search" loses the signal.
• Community presence is now a first-class AEO asset. Reddit (and forum content generally) is a 24% citation share on Perplexity and 5% on ChatGPT. An AEO program that ignores community channels is leaving a meaningful slice of both on the table.
• Protect head-term SEO, stop over-investing in mid-funnel informational. Your #1 for head terms still matters for brand defense. The mid-funnel informational pages your SEO team has been shipping since 2019 are the ones being cut by AIO. That effort needs to redirect.
• Get citation tracking in place now. Ahrefs Brand Radar, Parse, Profound, and HubSpot AEO Grader all launched or matured in early 2026. Without weekly visibility into who cites you in which engine, you can't run an AEO program. You're just hoping.

These implications are where the operational layer matters. A human SEO team tracking rankings in a spreadsheet cannot feasibly audit citations across ChatGPT, Perplexity, Gemini, and AI Overviews weekly for a 500-URL site. It is agent work: scheduled crawls, structured diffs, and automated triage.

Limitations

Three caveats worth naming explicitly.

First, the overlap studies use long-tail queries. Head-term behavior looks different: Google's cached answers and AI Overviews both lean more SERP-aligned on high-volume head terms. If your program is head-term-heavy, the 12% overlap number understates your SERP carry-through.

Second, citation data ages fast. The GPT-5.3 Instant shift in domain count happened inside a single model update. The 73% Reddit-on-ChatGPT growth was a three-month delta. Any AEO baseline built in April 2026 may not describe July 2026. Continuous measurement, not point-in-time audits.

Third, our operational observations are qualitative. We did not run a controlled experiment. Where Passionfruit pattern-observations are cited, we framed them as such. The headline number and the six findings all come from public third-party studies with stated sample sizes.

Ranking #1 on Google used to mean you'd been found. In 2026, it means you showed up to 12% of the room.

What we're watching next

Two follow-on questions this research surfaced, both worth their own studies.

• Does citation share predict revenue the way SERP share used to? Early signal in our client work says yes, but the attribution is messier than SEO analytics ever was. We're scoping a multi-quarter measurement.
• How quickly does new content get cited once published? Anecdotally the AIO-to-citation lag is shorter than SERP-to-rank, but we don't have a controlled sample yet. We're building one.

If you want to see how we run AEO monitoring and the agent workflows around it, join the waitlist. The underlying playbook is the same one we refined across Passionfruit's work with 500+ brands, now opened up as the product. For the operator essay on where this sits in the broader shift, see Agentic marketing vs marketing automation.