Speed-Date a Consultant

System Audit

In the News

Linkedin Youtube
Contact Us

Why Your Brand Doesn’t Show Up in AI Answers — Even Though You Rank Well on Google

Short answer: because AI answer engines don’t use Google’s ranking system, and multiple independent studies prove it. A 2026 SIGIR study measuring 11,500 queries found the sources retrieved by Google Search, AI Overviews, and Gemini overlap at less than 0.2 average Jaccard similarity. Said another way, the three systems, all owned by the same company, barely agree on what to cite. Your #1 ranking was earned in one of those systems. AI answers come from the others.

So let’s do this the way researchers would: what does the data actually show?

How do AI engines select sources compared to traditional search engines?

Grossman et al. (2026), in a study accepted to ACM SIGIR, the top information-retrieval conference, compared what Google Search, AI Overviews, and Gemini retrieve for the same 11,500 real-user queries. The retrieved sources were “substantially different for each search engine,” with less than 0.2 average Jaccard similarity between systems.

Xu, Iqbal & Montgomery (2026) ran 55,393 queries over 40 days and found that nearly 30% of domains cited in AI Overviews don’t appear anywhere in the first page of results shown alongside them. Their conclusion, verbatim in spirit: AIOs use “a source selection mechanism distinct from Google’s ranking algorithm.”

These findings indicate that Google’s own AI product ignores Google’s own rankings almost a third of the time. And the industry data shows the gap widening, not closing. Ahrefs’ July 2025 analysis found 76% of AI Overview citations came from Google’s top 10. Their March 2026 update, across 863,000 SERPs, found that number had collapsed to 38%. That’s a straight 50% haircut lesss than in a year (Search Engine Journal’s coverage puts the organic-only figure at 37%, with 36% of cited pages falling outside the top 100 entirely). And for AI assistants proper, Ahrefs’ earlier study found only 12% of AI-cited URLs rank in Google’s top 10 for the same query. (March 2026)

If good SEO equaled good GEO, that overlap should sit at 90%+. Consistently. Instead it’s 38% and falling for Google’s own AI product, and 12% for everyone else’s.

Ranking well on Google is evidence you optimized for one retrieval system. It is not evidence you’re visible to the others.

How important is third-party placement for AI visibility?

AI engines cite what other people say about you — at a rate of roughly 6 to 1

The single most important number for any brand asking this question comes from Zatuchin (2026), who analyzed 167,551 URL-grounded citations across 128 brands, 12 markets, and 13 languages. When an LLM answers a question about a company:

85.7% of citations point to sites the brand does not own. 14.3% point to owned properties.

Chen et al. (2025) reached the same conclusion experimentally, running controlled tests across verticals and languages: AI search exhibits “a systematic and overwhelming bias towards earned media”, third-party, authoritative sources, compared to Google’s more balanced mix of owned and earned results.

This is the mechanical explanation for the gap. Your Google ranking is substantially a function of what your site does — your pages, your keywords, your backlink profile. Your AI visibility is substantially a function of what everyone else’s sites say about you. A brand can dominate the first game and have essentially no position in the second, because they’re scored on different surfaces.

The citation supply is also brutally concentrated: Zatuchin found 80% of all brand citations come from roughly 18% of domains, following a Zipf distribution, with Wikipedia the single most-cited domain in 11 of 12 languages studied. There is a short list of sources the models trust. Either you exist on it, or you don’t exist.

Does domain authority help get cited in LLMs?

In a word: Nope. Vishwakarma, Kumar & Jamidar (2026) ran the cleanest experiment yet: 252,000 controlled trials across six LLMs, injecting two competing sources into the model’s context, varying exactly one content factor at a time, with brands anonymized and order counterbalanced. What made a source get cited first?

The biggest drivers: topical relevance and list position. Consistent secondary gains: explicit price/data points and recent timestamps. Unfortunately, things that had only small effects: trust cues (yikes!) and formatting-only edits.

That last part should recalibrate a lot of GEO advice. Retrofitting bullet points onto an off-topic page does nothing. What wins is content that is precisely on the question, positioned early, with concrete data and freshness signals. This is consistent with the foundational GEO paper (Aggarwal et al., KDD 2024), which found that adding statistics, quotations, and source citations boosted generative-engine visibility by up to 40%. Note that effective tactics vary significantly by domain.

Precision beats authority.

What technical and structural elements keep you from being mentioned by LLMs?

Grossman et al. also documented that websites blocking Google’s AI crawler (Google-Extended) are “significantly less likely to be retrieved by AIOs, despite having access to the content.” That’s a measured effect, not a hypothetical. A robots.txt line added in 2023 to keep AI from training on your content may be silently excluding you from AI answers in 2026.

The same study found AI Overviews are inconsistent between two runs of the same query and not robust to minor query rephrasing — and Chen et al. found engines differ significantly from each other in freshness, domain diversity, and phrasing sensitivity. Practically: your brand can appear in ChatGPT and not Perplexity, appear on Monday and not Tuesday. That’s not noise in your tracking tool. That’s the measured behavior of the systems.

The stakes are the answer box itself

There was a time when Google was rolling out AI Overviews for terms little by little. At this point, they’re pervasive. Xu et al. measured AI Overview activation at 13.7% of all trending queries, but 64.7% of question-form queries. Grossman et al., using representative real-user queries, measured 51.5%. The exact number depends on query mix, but the direction is unambiguous: the majority of question-shaped searches, the ones your buyers ask, now get a synthesized answer above the blue links. If you’re not in the synthesis, your #1 blue link is negotiating from underneath it.

What to do if you place in Google, but not in an LLM

There’s nothing worse than your CEO asking ChatGPT about your category and NOT seeing your company first. Or worse, not mentioned at all. So if you find yourself in a position of having great SEO, but falling behind in GEO, it’s time to act. Here are five things you can do right now:

  1. Rebalance toward earned and third-party surfaces. The 85.7/14.3 split is the map. Reviews, comparison content, trade press, Wikipedia-eligible presence, the community threads models demonstrably cite. In our GEO/AEO Jobs Tracker (122 postings, June 2026), 74% of GEO roles are scoped for on-site work only, companies are staffing against the 14.3% and ignoring the 85.7%.
  2. Publish precisely on the questions buyers ask. Topical relevance is the #1 measured citation driver. Question in the slug, in the H1, answered in the first two sentences. Look at this blog as the example.
  3. Load content with data, dates, and specifics. Statistics and citations: up to +40% visibility (Aggarwal et al.). Price info and recent timestamps: consistent measured gains (Vishwakarma et al.). Generic advice gets synthesized without you; unique numbers force attribution. Again, we did that here.
  4. Audit your crawler access today. Check robots.txt for GPTBot, Google-Extended, PerplexityBot, ClaudeBot. Blocking is a measured retrieval penalty.
  5. Track per-engine, not in aggregate. <0.2 Jaccard similarity between engines means there is no single “AI visibility.” Measure ChatGPT, Perplexity, and AIO citations separately, against the exact questions that drive your pipeline.

The “GEO is just SEO” crowd has a 4-million-person incentive to tell you your rankings have you covered. The research — six independent teams, controlled experiments, hundreds of thousands of measured citations — says the systems are different, the source bias is quantified, and the gap between ranking and being cited is structural, not temporary.

Your Google position is real. But in the world of changing buyer behaviors, it’s probably just not enough.

References

FAQ

Why does my brand rank #1 on Google but not appear in AI answers? Because AI engines use a source-selection mechanism distinct from Google’s ranking algorithm. Peer-reviewed measurement (Grossman et al., SIGIR 2026) found less than 0.2 Jaccard similarity between sources retrieved by Google Search, AI Overviews, and Gemini for the same queries, and ~30% of AI Overview citations don’t appear in first-page results at all (Xu et al., 2026).

Do AI models prefer third-party content over brand websites? Yes, overwhelmingly. Analysis of 167,551 brand citations found 85.7% point to sites the brand doesn’t own (Zatuchin, 2026), and controlled experiments show a systematic bias toward earned media over brand-owned content (Chen et al., 2025).

What actually makes AI engines cite a page? In 252,000 controlled trials across six LLMs, the strongest drivers were topical relevance and list position, followed by explicit data (like pricing) and recent timestamps. Formatting-only changes had little effect (Vishwakarma et al., 2026). Adding statistics, quotations, and citations boosted visibility up to 40% (Aggarwal et al., KDD 2024).

Can blocking AI crawlers hurt my AI visibility? Yes — measurably. Sites blocking Google’s AI crawler are significantly less likely to be retrieved by AI Overviews even when the content is otherwise accessible (Grossman et al., 2026).

Is GEO just SEO with a new name? No. The systems retrieve different sources (<0.2 overlap similarity), weight third-party consensus over owned content (85.7% vs 14.3%), and respond to different optimization levers. SEO remains necessary for Google; it is empirically insufficient for AI visibility.

  • Andrea Lechner- Becker

    AUTHOR

    Chief Strategy Officer at GNW Consulting

    Hard problems are Andrea’s favorite to solve. She believes solving big problems requires a forensic approach. Through systematic and scientific methods, all problems can be solutioned.