GEO Is Not a Second SEO Channel: What Actually Gets You Cited by ChatGPT, Perplexity and AI Overviews

Generative engine optimization (GEO) is the work of getting your business named and cited inside the answers that ChatGPT, Perplexity, Gemini and Google AI Overviews write. It is not a separate marketing channel you buy alongside SEO, and most of the tactics sold under the GEO label do not survive contact with the published evidence. What the research consistently points to is narrower and less exciting: third-party authority, and content a machine can extract a reason from.

Where the term came from, and what it actually means

GEO was named in a November 2023 paper by Pranjal Aggarwal, Vishvak Murahari and four co-authors, later accepted to KDD 2024. Their framing was practical. When a search engine answers in a synthesized paragraph rather than a list of links, the thing worth optimizing is no longer position one. It is whether your URL appears in the citation set behind that paragraph, and how much of the answer leans on it.

The paper built a benchmark of user queries across multiple domains and tested content changes against it. Their headline result was that these methods lifted visibility inside generative responses by up to 40%, with the largest gains going to sites that were not already ranking near the top. That last detail matters more than the 40% for most businesses reading this, and we will come back to it.

So the definition is simple. Traditional SEO competes for a slot on a results page. GEO competes for a mention inside an answer. The interesting question is how different the work required actually is.

How GEO differs from SEO, and where it does not

Start with the part that surprises people: Google's own position is that there is no separate discipline here at all. Its official guide on generative AI features addresses AEO and GEO by name and states that "optimizing for generative AI search is optimizing for the search experience", and therefore still SEO. Google's AI features pull from the same index, judged by the same ranking systems. A page must be indexed and eligible to show with a snippet before it can be cited at all.

That kills the idea of GEO as a parallel channel with its own budget line and its own index. But it does not mean nothing changed. Three things genuinely did:

• The unit of success moved from ranking to citation. A page can rank on page one and still be absent from the answer above it.
• Retrieval got broader. Google describes a query fan-out technique, where the system issues several related sub-queries and pulls sources for each. One page can now be reached through questions you never targeted.
• Non-Google engines source differently. This is where GEO stops overlapping with SEO, and it is the part most businesses have not accounted for.

The finding that should reshape your budget

A team at the University of Toronto ran a large controlled comparison of AI engines against Google across verticals, languages and query phrasings. Their central result was a "systematic and overwhelming bias towards Earned media" in AI search: independent reviews, publishers and comparison sites, rather than brand-owned pages or social posts.

The numbers are not marginal. For queries about well-known brands, ChatGPT's cited sources were 93.5% earned media and 6.5% brand-owned, with no social sources at all. For niche brands the tilt was sharper still at 95.1% earned. Claude sat around 87.3% earned. Perplexity was the most inclusive of the group, at 67.4% earned with 23.8% social, and Gemini the most balanced at 63.4% earned and 25.1% brand. Google, by contrast, kept a far more even mix of brand, earned and social across the same queries.

Read that as a spending instruction. If your entire content investment sits on pages you own, you are optimizing for the source type that ChatGPT uses least. The brands that get recommended are the ones written about elsewhere.

The same study found local service categories are the least aligned of all. Overlap between AI citations and Google results reached only 20.6% for home cleaning and fell to 0.1% for IT support. Ranking well locally on Google tells you almost nothing about whether an AI assistant will name you.

What Google says you can safely ignore

A large share of GEO advice sold today is addressed directly in Google's documentation, and dismissed. Its mythbusting section is unusually blunt for Google. For Google Search specifically:

• llms.txt and similar AI files. Google Search ignores them. Creating one neither helps nor harms your visibility.
• Chunking content into small pieces. Not required. Google states its systems handle multiple topics on a page and surface the relevant part. There is no ideal page length.
• Rewriting content specifically for AI systems. Not needed, because the systems understand synonyms and intent rather than requiring exact phrasing.
• Chasing inauthentic mentions. Google calls this less useful than it appears, and notes its spam systems are part of what generative features depend on.
• Treating structured data as an AI requirement. It is not required for generative features. Google still recommends keeping it for rich results, which is a different and still worthwhile reason. Our explainer on what schema markup does covers where it genuinely earns its place.

One more warning worth repeating: Google explicitly cautions against third-party tools claiming access to internal Google ranking or AI metrics. No such access exists.

What to actually do, in priority order

Strip out the noise and the evidence-backed work is short enough to fit on one page.

1. Earn third-party coverage before writing another owned page. Given that earned media dominates AI citations, review placements, industry publications, expert roundups and credible directory listings are the highest-leverage spend available. This is public relations doing SEO work.
2. Make your pages easy to extract a reason from. AI answers recommend a shortlist and justify each entry. Comparison tables, explicit pros and cons, plain statements of who a product suits and who it does not, and clear pricing and warranty facts all give a model something to quote. Vague positioning copy gives it nothing.
3. Fix eligibility first. Indexing, crawlability, snippet eligibility and a sound technical base are preconditions, not optimizations. A page that cannot be indexed cannot be cited, and one that was not ranking before has less chance now.
4. Cover the whole lifecycle, not just the top of the funnel. Gaps in post-purchase, troubleshooting and comparison content are where an AI reaches for a competitor instead.
5. Treat each engine as a separate surface. Perplexity rewards social and video presence that ChatGPT ignores entirely. Gemini leans further on brand-owned pages than either. A single tactic applied everywhere will underperform on most of them.
6. Measure with real data. Google's Generative AI performance report in Search Console shows your actual AI feature impressions and clicks. For the non-Google engines, run your own recurring prompt set and record who gets named.

How to tell if your current setup is failing

Four signs, in rough order of how often we see them:

• Your competitors appear in independent "best of" roundups for your category and you do not. On the earned-media evidence, this alone can decide the citation.
• Your service pages describe what you do but never state who you are the right fit for, or against what alternative.
• Rankings are stable but qualified enquiries are drifting down. A Pew Research field study, cited in the Toronto analysis, found link clicks fell to 8% when an AI summary was present compared with 15% without, with about 26% of those searches ending with no click at all. Our breakdown of what the 2026 AI Overviews data shows goes deeper into that gap.
• Nobody has ever checked. If you have not asked ChatGPT or Perplexity a buying question in your category and recorded who they name, you do not have a baseline.

The bottom line for a business owner

GEO is not a new channel, and it is not a bolt-on. It is a shift in what SEO is being measured against, plus a genuine change in emphasis toward third-party authority that most content budgets are not currently built for. The good news sits in the original research: the biggest visibility gains went to sites that were not already dominant. Being the smaller name in your category is less of a disadvantage here than it is on a traditional results page, provided the rest of the web has reason to talk about you.

Frequently Asked Questions

How long does it take to start getting cited by AI search engines?

Expect months rather than weeks, because the main driver is third-party coverage and that takes time to earn and to be recrawled. Technical eligibility fixes can register faster, sometimes within a crawl cycle, but they only make a page eligible. The slower work of getting written about elsewhere is what usually moves citation rates.

Do I need a separate GEO agency if I already have an SEO agency?

Not usually, and Google's own position is that this is still SEO rather than a separate discipline. The real question is whether your current agency is doing digital PR and earned-media work, structuring pages for comparison and justification, and tracking citations across non-Google engines. If the answer is no, you need those capabilities added, not a second vendor duplicating the first.

Can a small brand get cited when AI engines clearly favour big names?

Yes, and the research is mildly encouraging on this point. The original GEO study found the strongest visibility gains went to lower-ranked sites rather than incumbents. The constraint for smaller brands is earned coverage: AI engines cite independent sources heavily, so a niche brand with genuine review placements and expert mentions can outrank a larger one that has only its own website.

How much should a business budget for generative engine optimization?

There is no standard figure, and the honest answer is that it depends on three specific things: how much earned media you already have, how technically sound your site is, and how many engines you need to compete on. A business with strong existing press coverage may need only content restructuring. One starting from zero third-party mentions is funding a digital PR programme, which is the larger cost.

How do I measure whether GEO is working?

Use two measurements together. Google's Generative AI performance report in Search Console shows impressions and clicks from its AI features on your own pages. For ChatGPT, Perplexity and Gemini, build a fixed set of buying-intent prompts for your category, run them on a schedule, and record which brands and domains get named each time. Track share of mentions over time rather than any single result, since answers vary between runs.

If you want a read on where your business currently stands in AI answers, our generative engine optimization services begin with exactly that baseline, and our AI SEO services build the earned-media and content structure work on top of it. For a broader introduction to the topic, our earlier piece on what generative engine optimization means for digital marketing is a good starting point.