Updated
Updated What changed on
Fact-checked The GEO paper's results table and publication details were re-checked on arXiv on 14 October 2026, and the BrightEdge, Ahrefs, Google and AI crawler facts above were re-checked against the 14 October 2026 source checks. The Google search and ChatGPT message figures were re-checked on 14 October 2026 against Google's May 2025 blog post and OpenAI's NBER working paper. Nothing else on the page was re-checked.
- Corrected the GEO paper figure from 22-41% to 27-41%: that is the lift its four strongest rewriting methods gave a page's share of the AI answer.
- Corrected who wrote the paper: researchers from Princeton and IIT Delhi, published at KDD 2024.
- Removed the remaining credits of the GEO paper to Princeton alone, and the claim that it showed engines rewarding different levers. Its main tests used a GPT-3.5 engine; on Perplexity, adding quotations gave a 22% lift.
- Replaced the 13.1% AI Overview trigger rate with BrightEdge's February 2026 figure: about 48% of tracked queries, up from 30% a year earlier.
- Corrected the AI crawler lists: Google-Extended is a robots.txt token with no user agent, and the crawlers that matter for AI search are OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot.
- Replaced the Reddit claim with Ahrefs' September 2026 US data: reddit.com is the most-mentioned domain in ChatGPT answers, at 16.8%.
- Removed the claim that Perplexity favours fresh pages. Ahrefs found its in-text citations averaged 1,166 days old, older than ChatGPT's.
- Dropped FAQPage and HowTo from the schema list, as Google retired How-to rich results in 2023 and FAQ rich results on 7 May 2026, and noted that Google says structured data isn't required for its AI features.
- Replaced the undated "14 billion Google searches versus ChatGPT's 37 million" comparison with dated figures: Google's more than 5 trillion searches a year (May 2025) and more than 2.5 billion ChatGPT messages a day (July 2025), about a quarter of them seeking information.
SEO vs GEO: the honest difference.
The 30-second answer
SEO gets your page ranked so a user clicks. GEO gets your brand cited inside an AI engine's answer, often with no click. Roughly four-fifths of the underlying work is shared - schema markup, structurally clean content, a real byline, technical health, authority. The new fifth is real: bot-access management for AI crawlers, citation-share tracking, sustained Reddit and forum posting, and the GEO-specific content levers the GEO paper (Princeton, IIT Delhi and others, KDD 2024) measured (cited sources, statistics and fluent wording each lifted a page's share of the AI answer by 27% to 31%, and adding quotations by 41%). You need both. The flat "GEO is repackaged SEO" line undersells that 20%; the "GEO is a whole new discipline" line oversells it.
Google
Bing
GEO wins citations on:
Perplexity
ChatGPT
AI Overviews
Gemini
Claude
Where SEO and GEO have measurably diverged
of URLs cited in Google AI Overviews appeared in the first 10 results blocks for that query. Ahrefs counts blocks, not organic results, so this is not the same as ranking in the top 10.
How it was measured: 863,000 keyword SERPs and 4 million AI Overview URLs. Ahrefs measures the top 10 as blocks, counting ads, featured snippets and video packs as separate blocks.
was Ahrefs' July 2025 figure. It counted only the top 3 citations per AI Overview, and Ahrefs has since changed its parsing, so the two years are not like for like.
How it was measured: 1.9 million citations from 1 million AI Overviews, analysing only the top 3 most visible citations in each one. Ahrefs notes it improved its parsing between this study and the 2026 update. Both differences mean the fall to 37.9% cannot be read as a like for like change.
of links cited by ChatGPT, Gemini and Copilot appear in Google's top 10 for the same prompt. Perplexity is the outlier at closer to one in three.
How it was measured: 15,000 prompts. This is 2025 data and has not been re-run, so treat it as directional rather than current.
The numbers that frame the decision
Three statistics most "SEO vs GEO" guides skip. They decide whether you treat this as a 2026 priority or a 2027 problem:
Side-by-side, the way I'd brief a board
| SEO | GEO | |
|---|---|---|
| Goal | Rank in search engine results | Be cited inside AI engine answers |
| Primary surface | Google SERP, Bing SERP | OpenAI's ChatGPT, Perplexity, Anthropic's Claude and Google's Gemini, plus the AI Overview block |
| Input signal | Keywords - short, head-term phrases | Prompts - long, conversational, often question-form |
| Measurement | Rank, click-through, organic traffic, conversions from search | Citation frequency, share-of-voice inside the generated response, branded mentions per prompt cohort |
| Core levers | Crawlability, indexability, content depth, internal linking, authority. Schema is hygiene, not a lever. | All of the above plus search crawler access (OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot), Reddit presence, named-author signals, and the GEO paper's content levers. Google says structured data isn't required for its AI features, and Ahrefs' May 2026 test found no citation lift from adding it. |
| Click impact | You get a visitor when you rank | You may get cited without a visitor - the user reads the answer in the AI engine |
| Freshness sensitivity | Evergreen content can rank for years | LLMs favour recency more aggressively - quarterly content refresh is the baseline |
| Toolchain | Ahrefs, Semrush, Screaming Frog, Search Console | Profound, AIClicks, Otterly, Semrush AI Visibility, plus the SEO stack |
| Maturity | 25+ year discipline | ~3 year discipline (2023 onwards) |
What "ranked" looks like vs what "cited" looks like
Most theoretical comparisons skip the visual that actually clarifies this. The same buyer question hits two different output shapes:
Both queries come from the same buyer doing the same job. The optimisation work to win both is mostly the same. The measurement and what the buyer does afterwards are not.
What is genuinely new in GEO
Strip the marketing and the actually-new work is narrow. The GEO walkthrough on this site takes each of these items and shows how to put it into practice:
- Citation-share tracking. You cannot measure rank for an AI engine the way you measure rank for Google. Tools like Profound, AIClicks and Otterly emerged in 2024-25 because no traditional SEO platform could see citation share-of-voice. Semrush's AI Visibility Toolkit is the established-platform answer to the same gap.
- Sustained Reddit and forum participation. ChatGPT draws on Reddit constantly. Ahrefs' September 2026 US data puts reddit.com as the most-mentioned domain in ChatGPT answers, at 16.8%. Sustained Reddit posting under a real account is a GEO tactic SEO does not really have an equivalent of.
- AI bot access management. OAI-SearchBot, GPTBot, Claude-SearchBot, ClaudeBot, PerplexityBot and Applebot each have their own user agent; Google-Extended is a robots.txt token with no user agent of its own. GPTBot, ClaudeBot and Google-Extended only control training: blocking them does not remove you from ChatGPT search, Claude search or AI Overviews. Managing this lot is a 2026 GEO line item; traditional SEO had only Googlebot to think about.
- Entity establishment for AI extraction. Wikipedia, Wikidata, Crunchbase and LinkedIn carry more weight for GEO than for traditional ranking. AI engines lean on these off-site signals to decide who is credible because they cannot run the same link-graph authority calculation Google does on a five-second timescale.
- Engine-specific content tuning. The GEO paper's levers (cited sources, statistics, fluency, quotations) were tested mainly on a GPT-3.5 engine summarising five Google results; in its one real-engine test, on Perplexity, adding quotations gave a 22% lift. The paper did not test ChatGPT or Google AI Overviews.
What is genuinely shared
- Schema.org markup, especially Article, Product, Person and Organization types (Google retired How-to rich results in 2023 and FAQ rich results on 7 May 2026)
- Page depth, clean structure, answer-first paragraphs near the top
- Real named byline plus the E-E-A-T signals (bio, credentials, sameAs into Wikipedia or LinkedIn)
- Site architecture, crawlability, indexability
- Authority signals - links, mentions, citations from credible sources
- Local presence (Google Business Profile, NAP consistency) - still matters for both surfaces
The engines do not behave the same way
"GEO" is a single label hiding five engines with different retrieval and ranking logic. A serious programme tracks each separately:
ChatGPT
Reddit and Wikipedia are the two most-mentioned domains in its answers (Ahrefs, US, September 2026). Synthesises answers, often without inline citations. Rewards direct, quotable phrasing and crisp definitions.
Perplexity
Most citation-transparent of the four. Surfaces sources by default. Rewards citation density and pages that read like research notes rather than marketing copy. Ahrefs found its in-text citations averaged 1,166 days old, older than ChatGPT's. Its data does not show Perplexity favouring fresh pages more than other assistants.
Google AI Overviews
Tracks classic Google ranking signals most closely. Featured-snippet logic carries over. The cheapest engine to win for a site that already does SEO well.
Gemini
Sits closest to the Google organic index. Less structured outputs than ChatGPT, fewer visible citations than Perplexity. Tends to follow the same patterns that work for AI Overviews.
The bit nobody says clearly: most "GEO services" being sold in 2026 are good SEO done with AI extraction in mind, plus citation tracking, plus a thin layer of GEO-specific content craft (the GEO paper's levers above). The SEO part does 70-80% of the lift. The GEO craft layer is real and has peer-reviewed evidence behind it - but pure GEO-without-SEO is mostly snake oil. If your current SEO partner has not retooled for AI search by 2026, replace them rather than buying a second adviser on top.
How to decide where to invest first
Three signals decide the order. Run through them honestly before committing budget:
- If your SEO foundations are broken - thin content, no schema, anonymous brand voice, technical health amber - invest in SEO first. GEO has nothing to optimise on a site that does not pass the basics. The GEO paper's levers do not work on a page Google cannot crawl.
- If your SEO foundations are solid but click-through is dropping while you still rank, AI Overviews are eating your click. That is the signal to invest in GEO at the page level - shorten the lead, add the citation density, push the named author harder, structure the answer-first paragraphs.
- If your category is among the roughly 48% of tracked queries that showed AI Overviews in BrightEdge's February 2026 data and your buyer demographic skews younger or technical, GEO has already become the leading indicator. Citation-share tooling plus dedicated Reddit posting cadence become first-tier budget items, not experiments.
What this means for hiring
The hiring question most directors of marketing get wrong: "Do I need to hire a GEO consultant?" The right framing is whether your existing SEO function has adapted. If they still talk about backlinks and keyword density and have nothing to say about citation share-of-voice or AI bot access, the answer is to replace them rather than add a parallel hire. If they already track AI citations and have a position on Reddit posting cadence, you have a GEO function already - call it whatever you want.
More on the SEO and GEO question.
Is GEO replacing SEO?
No. GEO is an additional surface, not a replacement. Google said in May 2025 that it sees more than 5 trillion searches a year, roughly 14 billion a day. OpenAI's own research put ChatGPT at more than 2.5 billion messages a day in July 2025, and only about a quarter of those were classed as seeking information. Across most UK B2B verticals during 2026, organic Google drives more pipeline than AI search does. The brands that win invest in both, weighted by their buyer mix.
Do I need a separate GEO budget?
Mostly no. Roughly four-fifths of GEO work folds into an existing SEO budget because the underlying skills overlap. The new line items are dedicated citation tracking platforms (Profound, AIClicks, Otterly), sustained Reddit and forum participation, and the AI-specific content levers documented in the GEO paper (Princeton, IIT Delhi and others, KDD 2024).
Should I learn SEO or GEO first?
SEO. GEO assumes the SEO fundamentals already work - schema markup in place, the page structurally clean, a real byline, the E-E-A-T signals consistent. Without those, no amount of GEO work has anything to optimise. Learn SEO first, then GEO becomes the extra fifth.
Is GEO a real discipline or repackaged SEO?
Both, depending on the practitioner. Roughly four-fifths of good GEO work is good SEO done with AI extraction in mind. The remaining fifth is genuinely new - AI engine citation tracking, sustained Reddit posting, bot access management for crawlers that did not exist three years ago, and the GEO-specific content levers (cited sources, statistics, fluency, quotations) the GEO paper measured.
Do I need a separate GEO consultant?
If your current SEO partner has not retooled for AI search by 2026, replace them rather than paying two. The disciplines share most of the work. The one exception: very large enterprises running parallel teams may benefit from a dedicated AI search lead reporting into the same SEO function.
Does the same GEO work apply to ChatGPT, Perplexity and Gemini?
Mostly, but not entirely. The shared signals (schema markup, clean structure, real byline, citation density) lift visibility across all four major engines. The engine-specific differences matter at the edges - Perplexity cites heavily and visibly (though Ahrefs found its in-text citations averaged 1,166 days old, older than ChatGPT's), ChatGPT leans on Reddit and Wikipedia, Google AI Overviews tracks classic Google ranking signals closely, and Gemini sits closest to Google's organic index. A serious GEO programme tracks visibility separately per engine rather than treating them as one channel.
Jason Burns
Independent UK SEO, GEO and AI consultant. In search since 2007. Portfolio includes 3M, BlackRock, Unilever and E.ON. Owner of SEO Moves Ltd since 2014.
More about Jason →Not sure if you need SEO or GEO investment first?
Send your site URL. I come back with an honest call - SEO first (if the foundations need work), GEO first (if foundations are solid but click-through is dropping), or both running in parallel.
Send your URL