By 
Shreyas M V · Published · 14 min read
GEO Without the Hype: What Google Says, What Research Shows, What Is Still Guesswork
Generative engine optimization (GEO) is the work of getting a site into the answers AI search systems write, and Google's own guide says that for Google Search it is still SEO. The strongest evidence backs the SEO basics: indexable pages, original content and accurate markup. Quotes and statistics have one reported study behind them, and most other advice is practitioner opinion. This page grades each popular tactic by its evidence and lists what Google says you can skip.
Key points
- Google says optimizing for generative AI search is still SEO, because its AI features rely on its core ranking and quality systems.
- Nine tactics are graded below as official guidance. Most of the rest are practitioner opinion, and three are unknown.
- Google says to skip or avoid at least nine tactics, including separate pages for every query variation and llms.txt files.
- The only quantified lift, 30% to 40% for quotes and statistics, comes from an unnamed study reported by Semrush.
- Guidance from Google covers Google. No source read here documents how ChatGPT, Perplexity or Claude choose what to cite.
What is generative engine optimization, and how does AEO relate to it?
Generative engine optimization is the work of making a site or brand more likely to appear in the answers that AI search systems write.
The term comes from a 2023 paper by Pranjal Aggarwal and co-authors, later presented at the KDD 2024 conference (arXiv:2311.09735), according to Wikipedia's entry on the topic. The same entry says usage varies across practitioners, vendors and publications and that there is no consensus definition.
Answer engine optimization (AEO), large language model optimization (LLMO), AI optimization (AIO) and AI SEO are listed as names for the same idea. The table shows who uses which label.
| Who | Term used | Position |
|---|---|---|
| Says "AEO" and "GEO" are terms you may see used | Optimizing for generative AI search is optimizing for the search experience, so it is still SEO. Wikipedia dates the guide 10 July 2026. | |
| Microsoft Bing | GEO | Uses the term in its documentation and says SEO fundamentals support eligibility for AI-generated experiences (Search Engine Journal, 27 February 2026). |
| Microsoft Advertising | AEO and GEO as separate things | Defines AEO as optimizing for AI agents and assistants such as Copilot, and GEO as optimizing for generative AI search environments (Search Engine Roundtable, 14 January 2026). |
| HubSpot | AEO | Built answer engine optimization into its platform (TechTarget, 14 April 2026). |
| Profound | AEO | Says AEO and GEO are the same practice and prefers AEO because GEO can be confused with geography (June 2025). |
| Perplexity | GEO | A representative used the term while cautioning against vendors overstating their expertise (Search Engine Roundtable, 19 November 2025). |
| Forrester | AEO and related terms | Nikhil Lai argued in 2025 that they are "significantly, but not fundamentally, different from SEO". |
Kanso's reading is that the labels overlap heavily. Judge any advice by the evidence behind it, not by whether it says GEO or AEO.
What does Google say about optimizing for AI answers?
Google says optimizing for generative AI search is optimizing for the search experience, and so it is still SEO. Its guide to generative AI features gives the reason: AI Overviews and AI Mode are rooted in the core Search ranking and quality systems.
The guide names two techniques behind them. Retrieval-augmented generation grounds an answer in pages that core ranking systems retrieve from the Search index, and query fan-out runs several related searches at once. Google's example is a question about fixing a lawn full of weeds, which can fan out into searches on herbicides, chemical-free removal and prevention.
The guide's date is not settled in the sources read. Wikipedia's reference list dates it 10 July 2026 and also cites Search Engine Journal coverage from 15 May 2026 and 7 June 2026, and Wikipedia adds that Google had voiced the position at conferences before 2026 and had not put it in official documentation. Treat “July 2026” as approximate and check the page's own date before quoting it.
Google also gives a test for any decision: ask whether visitors would find the content satisfying. It says unique, useful content will likely influence presence in generative AI search more than any other suggestion in the guide.
One limit applies to everything Google says here. The guide covers Google's own features, so every “official guidance” grade below applies to Google, or to Bing where stated, and no further.
How are GEO tactics graded on this page?
Each tactic gets one of four grades, set by the strongest kind of source behind it. The grades describe the evidence, not whether the tactic works.
- Official guidance: stated in documentation from the search engine the tactic targets. It applies to that engine only.
- Research: a named study with a stated method. Here, every research item reaches Kanso second-hand, and that limit is stated on each.
- Practitioner opinion: advice from a vendor, publisher or analyst with no stated method, including hedged claims that something “seems” or “might” work.
- Unknown: no source read for this page supports or contradicts the tactic.
How the tables were built: Kanso collected the tactics that appear in Google's documentation and in the three top-ranking GEO guides (Semrush, Wikipedia and Coursera), then graded each by its strongest source, read on 21 October 2026. The first table holds the tactics with official support.
| Tactic | Grade | Evidence and limits |
|---|---|---|
| Make the page crawlable, indexed and eligible to show a snippet | Official guidance (Google) | Google's guide says a page must be indexed and snippet-eligible to appear in generative AI features. Its May 2025 post lists the basics: Googlebot not blocked, an HTTP 200 status and indexable content. |
| Publish non-commodity content with a point of view the web does not already have | Official guidance (Google) | Google's guide says this will likely influence presence in generative AI search more than any other suggestion in it, and gives first-hand experience as the example. |
| Add high-quality images and video where they help | Official guidance (Google) | Google says its generative AI features can bring in images and video, which gives a page more ways to appear than a text link. |
| Give visitors a good page experience | Official guidance (Google) | Google asks for a page that displays well on every device, loads quickly and makes the main content easy to tell apart from everything else, for visitors from classic and AI results. |
| Keep structured data matching the visible page | Official guidance (Google) | Google says markup must follow its guidelines and everything in it must also be visible on the page. The text read does not claim markup raises inclusion in AI features. |
| Set snippet and indexing controls on purpose | Official guidance (Google) | Google says nosnippet, data-nosnippet, max-snippet and noindex are the controls, and that more restrictive settings limit how your content is featured in its AI experiences. |
| Follow JavaScript SEO basics if the site relies on scripts | Official guidance (Google) | Google says it can process JavaScript content as long as it is not blocked, though sites built on JavaScript frameworks are generally harder to optimize. This covers Google only. |
| Judge visits by conversions, not clicks alone | Official guidance (Google) | Google says clicks from results with AI Overviews are higher quality and suggests tracking sales, signups and similar outcomes. That is Google's claim, and the post read publishes no data for it. |
| Keep SEO fundamentals sound for Bing | Official guidance (Microsoft Bing) | Wikipedia reports that Bing's documentation says SEO fundamentals support eligibility for AI-generated experiences. Kanso did not read Bing's documentation directly. |
The second table holds everything else. None of these rows has official guidance behind it, and the evidence is thinner the further down it goes.
| Tactic | Grade | Evidence and limits |
|---|---|---|
| Add quotes and statistics to content | Research (reported second-hand) | Semrush reports a study of 10,000 real-world queries in which pages with quotes and statistics had 30% to 40% higher visibility in AI responses. The text read does not name the study or define visibility, so the figure is unchecked. |
| Earn unlinked brand mentions across the web | Practitioner opinion | Semrush says unlinked mentions seem to carry more weight with AI systems. It offers no study for the claim. |
| Render content on the server | Practitioner opinion | Semrush says AI crawlers seem to struggle with JavaScript, so script-dependent content might be invisible to them. No test is cited, and Google says it processes JavaScript. |
| Keep content fresh | Practitioner opinion | Semrush says AI tools often favor fresh content, and Coursera says outdated or surface-level pages may lose out. Neither cites a test. |
| Have an accurate Wikipedia entry | Practitioner opinion | Semrush says Wikipedia seems to be a large share of AI training data, so an entry could raise the odds of a mention. No data is given. |
| Build a presence on Reddit, YouTube and similar platforms | Practitioner opinion | Semrush says platforms with user-generated content appear to have high exposure in generative engines. Wikipedia's reference list includes a 31 March 2026 Search Engine Land study headline that AI search engines cite Reddit, YouTube and LinkedIn most. Kanso did not read the study, so its method is unchecked. |
| Write long-tail, question-style keywords in a conversational tone | Practitioner opinion | Coursera recommends it. Google observes that people ask longer, more specific questions in AI search, but does not tell sites to change their writing style for AI. |
| Answer related questions on the same page | Practitioner opinion | Coursera recommends it. Google supports serving what users want, and warns against separate pages built for every query variation or fan-out query. |
| Use an AI visibility tool to steer the work | Unknown | Wikipedia says the efficacy of third-party tools is debated and cites a Digiday report of 1 May 2026 on marketers questioning costly tools because of inconsistent results. Kanso did not read that report. |
| Expect one set of tactics to work on ChatGPT, Perplexity, Claude and Gemini | Unknown | Google's and Bing's documentation covers their own products. No source read for this page documents how the other assistants choose what to cite, and Coursera notes that AI models do not disclose how they do it. |
| Hire a GEO service to raise AI citations | Unknown | No source read reports a controlled result for a GEO service. Perplexity's representative warned about vendors overstating their expertise (Search Engine Roundtable, 19 November 2025, per Wikipedia). |
Which GEO tactics does Google say to skip or avoid?
Google's guidance says to skip or avoid at least nine tactics, four of them stated in the text read for this page and five recorded from an earlier reading. Google's guide has a section headed “Mythbusting generative AI search: what you don't need to do”, and the extract Kanso used for this page stops before it.
The last five rows therefore come from Kanso's weekly growth checklist, which recorded the guide on 13 October 2026. Confirm them against the live guide before relying on them.
| Tactic | What Google says | Where recorded |
|---|---|---|
| Separate pages for every query variation or fan-out query | Doing so primarily to manipulate rankings or AI responses violates the scaled content abuse spam policy. A high quantity of pages does not make a site higher quality. | Google's AI optimization guide, read 21 October 2026 |
| Generating many pages with AI without adding value | It may violate the scaled content abuse policy. Fact-check all AI-generated content, including titles, meta descriptions, structured data and image alt text. | Google's guidance on generative AI content (page shows an update of 1 October 2026) |
| Recycling what the web already says | A summary of existing content restates information available elsewhere. Google asks for a unique point of view and content a generative AI model could not easily produce. | Google's AI optimization guide, read 21 October 2026 |
| Chasing perfect semantic HTML for AI | It is not required. Write for human readability, though semantic HTML still helps screen readers. | Google's AI optimization guide, read 21 October 2026 |
| llms.txt and other AI text files | Google Search does not use them. | Kanso's weekly growth checklist, recording a reading of the guide on 13 October 2026 |
| Chunking content into tiny pieces | Not required. | Kanso's weekly growth checklist, same reading |
| Rewriting content just for AI systems | Not required. | Kanso's weekly growth checklist, same reading |
| Seeking inauthentic mentions | Not as helpful as it might seem. | Kanso's weekly growth checklist, same reading |
| Special markup for AI features | Not required. Structured data is still worth using for rich results. | Kanso's weekly growth checklist, same reading |
Kanso's reading is that the list has one pattern: Google discourages work aimed at AI systems separately from readers. The tactics it supports are the ones that serve a person who lands on the page.
What does the research show about GEO?
The research base is thin: one academic paper that coined the term, one reported study on quotes and statistics, and click-through figures that sources read differently. The extracts read for this page describe the paper's origin but not its method or results, so Kanso does not summarize its findings.
Semrush reports one study of 10,000 real-world queries that found pages with quotes and statistics had 30% to 40% higher visibility in AI responses than pages without them. The text does not say who ran the study, what counted as visibility or which assistants were tested. Kanso cannot confirm it is the same work as the Aggarwal paper.
On clicks, Coursera cites an Ahrefs analysis, accessed on 20 October 2025, finding a 34.5% lower average click-through rate for the top results when an AI Overview appears, compared with similar informational searches without one. As arithmetic only: a position that earned 100 clicks without an Overview would earn 100 × (1 − 0.345) = 65.5 with one, if the average applied to it. That is not a forecast for any site.
Google makes a different claim. Its May 2024 announcement says links inside AI Overviews get more clicks than the same page would get as a traditional listing for that query, and its May 2025 post says clicks from results with AI Overviews are higher quality. Neither post publishes data.
Kanso's reading is that the two claims measure different things. Ahrefs compares results pages with and without an Overview, while Google compares a link inside an Overview with the same link as a plain listing. Later figures from Ahrefs and Pew Research Center are in Kanso's weekly growth checklist.
What is still guesswork in GEO?
Four things are still guesswork, and each one is a place where paid advice is most likely to outrun evidence.
- How assistants other than Google pick sources. Coursera lists the lack of transparency as a challenge, and no primary documentation from those assistants was in the sources read.
- Whether mentions, rendering and freshness change AI inclusion. Semrush hedges each of these with “seems”, “might” or “could”, and cites no test.
- Whether visibility tools measure anything stable. Wikipedia says tool efficacy is debated, and cites a Digiday report of 1 May 2026 on inconsistent results.
- Whether any service can reliably cause citations. No source read reports a controlled result.
Coursera adds that AI systems change frequently, so a tactic that works this quarter may not work next quarter. Kanso's suggestion is to treat these as hypotheses: test them on one page at low cost, log the date and the assistant, and do not put a budget behind them.
How does GEO differ from SEO?
For Google Search, GEO is SEO. For other assistants the difference lies mostly in what you measure and where you earn mentions, and sources disagree on how large the gap is.
| Dimension | SEO | GEO |
|---|---|---|
| Goal | Rank in search results | Be included in AI-generated answers |
| Typical tactics | Crawlability, keywords, intent match, backlinks | Clear writing that is easy to extract, mentions on credible sites, fresh content |
| Typical metrics | Keyword rankings and organic traffic | AI visibility, mentions, citations and share of voice |
Even Semrush says much of the SEO work you already do carries over, and that treating the two as separate strategies might not make sense. Coursera says GEO builds on the same principles as SEO. Forrester's Nikhil Lai, per Wikipedia, argued in 2025 that advocates of the new terms tend to exaggerate the differences.
So is GEO replacing SEO? No source read says so: Semrush says it complements SEO, and Google treats the two as one. Kanso's reading is to treat GEO as a measurement and distribution layer on top of SEO, not a separate budget line.
How do you rank in Google's AI Overviews?
You do it with the same requirements that apply to Google Search, because Google says meeting its technical requirements covers you for search generally, including AI formats. It publishes no AI Overview-only checklist in the guidance read for this page, and it does not claim any step guarantees inclusion.
- Let Google find the page. Do not block Googlebot, return an HTTP 200 and keep the content indexable. If pages are crawled but not indexed, find which stage is broken first.
- Stay snippet-eligible. Google says a page must be eligible to show a snippet, and that restrictive snippet settings limit how content is featured in AI experiences.
- Write non-commodity content. Add what the average result lacks: first-hand experience, original analysis, a point of view.
- Support the text with images and video where they help, and keep Merchant Center and Business Profile details current for shopping and local queries.
- Give a good page experience on every device.
- Match structured data to the visible page.
- Measure in Search Console and judge visits by conversions, not clicks alone.
If an AI tool helped draft the page, Google says to fact-check all of it, including the title, meta description, structured data and image alt text, because generative models predict likely wording and may state things that are wrong.
How do you measure GEO performance?
Measure with the free first-party reports first, and treat third-party AI visibility scores as unverified until the vendor shows its method. Wikipedia names two free tools: the Search generative AI performance reports in Google Search Console and the AI Performance report in Bing Webmaster Tools.
Google's guide has a section on measuring visibility in generative AI features with Search Console, and Kanso's guide to that report covers what it shows and leaves out. For the wider Search Console routine, see how to read Google Search Console as a founder.
- Impressions in AI features from the Search Console report, as a trend over dates.
- Conversions from search visits, since Google advises looking at sales, signups and similar outcomes.
- A dated prompt log if you check assistants by hand: the prompt, the assistant, the date and whether your brand appeared. This one is Kanso's suggestion, and it records one person's checks, not a measurement.
Attribution is a practical limit. When an assistant names a brand without linking, there may be no visit in your analytics to count, so a mention and a session are different things.
Is GEO worth paying for, or just a buzzword?
The evidence supports GEO as a label for real work that is mostly SEO. It does not yet support paying for GEO as a separate service without seeing the vendor's method.
Money is moving into the category. Wikipedia's reference list records Profound raising $96 million (Fortune, 24 February 2026), Sitecore reportedly agreeing to acquire Scrunch for $225 million (Bloomberg, 3 June 2026) and a planned $1.9 billion Adobe acquisition of Semrush (TechTarget, 20 November 2025). Investment shows vendors expect demand, which is not the same as evidence that the tactics work.
The table is a framework built by Kanso from the sources named in its last column. Put each question to a vendor, or to your own plan, before spending.
| Check | A passing answer | Basis |
|---|---|---|
| What does the vendor measure? | Named assistants, a fixed list of prompts, dates and repeated runs, with the method shown | Wikipedia reports debate over tool efficacy and inconsistent results; Coursera says AI models do not disclose how they pick sources |
| Can the work be told apart from SEO? | Each task maps to indexing, content, markup or measurement that Google documents | Google's guide says this work is still SEO |
| Does any task risk Google's spam policies? | No pages built per query variation, and no AI-generated pages that nobody reviewed | Google's AI optimization guide and its generative AI content guidance |
| Is the claimed lift backed by a named study? | The study is named, the method can be read and the claim fits this article's grades | Semrush's 30% to 40% figure names no study in the text read |
| Who profits from the claim? | The vendor discloses whether it also sells the measurement tool | Semrush's guide recommends Semrush products; a Perplexity representative warned about overstated expertise |
A vendor that fails the first two checks is selling measurement or SEO under a new name. That can still be useful, but the price should reflect it.
What is documented, and what is Kanso's analysis?
The facts below come from named sources as of 21 October 2026, and the interpretations are Kanso's own.
- Documented: Google says its AI features rely on core Search ranking and quality systems, and calls optimizing for them SEO.
- Documented: a page must be indexed and snippet-eligible to appear in Google's generative AI features.
- Documented: Google says separate pages for every query variation, made to manipulate rankings or AI responses, violate its scaled content abuse policy.
- Documented: Semrush reports a 30% to 40% visibility lift for quotes and statistics from an unnamed study.
- Documented: Coursera cites Ahrefs for a 34.5% lower average click-through rate when an AI Overview appears.
- Documented: Wikipedia records no consensus definition of GEO and debate over third-party tools.
- Analysis: the GEO and AEO labels overlap, so judge advice by its evidence.
- Analysis: the Ahrefs and Google click claims measure different things and do not directly conflict.
- Analysis: tactics without official or method-backed support are hypotheses to test cheaply, not budget lines.
- Analysis: vendor funding is not evidence that a tactic works.
- Analysis: Kanso ran no tests and holds no first-party data on any tactic on this page.
Where does Kanso's GEO Agent fit?
Kanso's GEO Agent checks how readable your site is to AI systems, estimates how citable it looks, and turns the gaps into actions. In this article's terms it works on the technical and readability side, not on the rows graded unknown.
Its citability view is a model estimate. Kanso does not query ChatGPT, Gemini, Claude or Perplexity, so it is not a measurement of AI citations, and Kanso does not promise rankings or inclusion in any AI answer.
Generative engine optimization: frequently asked questions
What is generative engine optimization?
Generative engine optimization (GEO) is the work of getting a site or brand included in answers written by AI search systems. The term comes from a 2023 paper by Pranjal Aggarwal and co-authors, presented at KDD 2024, and Wikipedia records that there is no consensus definition.
What is the difference between GEO and SEO?
For Google Search there is no separate discipline: Google says optimizing for generative AI search is optimizing for the search experience, which it still calls SEO. Semrush frames GEO as aiming at inclusion in AI answers rather than ranking, while saying much SEO work carries over.
Is GEO replacing SEO?
No source read for this page says GEO replaces SEO. Semrush says GEO complements SEO, Coursera says it builds on the same principles, and Google treats the two as one.
How do you rank in AI Overviews?
Meet the same technical and quality requirements that apply to Google Search. Google says a page must be indexed and snippet-eligible, and that meeting the Search technical requirements covers you for search generally, including AI formats. It also says unique, non-commodity content will likely matter more than any other suggestion in its guide.
What is AEO?
AEO stands for answer engine optimization, and in most sources it is another name for GEO. Profound calls them the same practice, while Microsoft Advertising defines AEO as optimizing for AI agents and assistants and GEO as optimizing for generative AI search environments.
Is GEO the new SEO or just a buzzword?
It is a new label for work that is mostly SEO, with a few unproven additions. Google and Forrester both argue the differences are overstated, and the tactics that are not SEO rest on practitioner opinion rather than published tests.
Will GEO as a service work?
The evidence read for this page does not show that it does. No source reports a controlled result for a GEO service, so ask any vendor to show its method and whether its tasks go beyond SEO fundamentals before paying.
How do you measure GEO performance?
Start with the free first-party reports: Wikipedia lists the Search generative AI performance reports in Google Search Console and the AI Performance report in Bing Webmaster Tools. Google also advises tracking conversions such as sales and signups, not clicks alone.
Are there risks to generative engine optimization?
Yes. Coursera names three: AI models do not disclose how they pick sources, systems change often, and summaries can distort facts. Google adds a policy risk, since mass-producing pages for query variations or generating pages without added value may violate its scaled content abuse policy.
Do quotes and statistics help content appear in AI answers?
One study reported by Semrush found 30% to 40% higher visibility for pages with quotes and statistics, but the study is not named in the text read and visibility is not defined. Treat it as a reason to test, not as established fact.
Sources
- Optimizing your website for generative AI features on Google Search, Google Search Central
- Top ways to ensure your content performs well in Google's AI experiences on Search, Google Search Central Blog, May 2025
- Google Search's guidance on using generative AI content on your website, Google Search Central
- Generative AI in Search: Let Google do the searching for you, Google, The Keyword, May 2024
- Generative engine optimization, Wikipedia
- Generative Engine Optimization: A Practical Guide, Semrush
- What Is Generative Engine Optimization?, Coursera
Note: Facts are attributed to Google documentation, Wikipedia (a secondary source, whose own references Kanso did not open) and the named vendors' pages, and were checked on 21 October 2026. Kanso ran no tests of its own, so the grades describe the evidence, not whether a tactic works. Search rules change, so confirm against the linked source before acting. Kanso has no affiliation with Google, Microsoft, Semrush or Coursera.
Search & AI discovery