GEO vs AEO vs SEO: what each means and which to work on first
SEO, AEO and GEO name overlapping work, and Google counts all three as SEO. What each term means, what the KDD 2024 GEO paper and the NeurIPS 2025 C-SEO Bench measured, which reports show each one, and where a small team should start.

TL;DR
- SEO is work to rank in search results. AEO (answer engine optimization) and GEO (generative engine optimization) both name work to be the source an AI answer draws on. Google says optimizing for its AI features is still SEO, so start with SEO.
- The 2023 GEO paper found citations, quotations and statistics raised a page's share of a simulated AI answer by 30 to 40%. C-SEO Bench (2025), testing four models, found none of those rewrites raised citation rank significantly, and statistics lowered it in 19 of 24 settings.
- In C-SEO Bench, moving a document to the top of what the model read moved it up the citations more than any rewrite did. That is retrieval, which is ordinary SEO: crawlable, indexed, eligible for a snippet and ranking for the question.
- Measure with first-party data: Search Console's Performance and generative AI reports for Google, and Bing's AI Performance report for Copilot, Bing's AI summaries and select partner integrations. Check ChatGPT and Perplexity citations by hand.
What SEO, AEO and GEO each mean
SEO, AEO and GEO are three names for the work of being found through search, and they differ in which result they aim at. SEO, search engine optimization, is the work of earning a place in a search engine's ranked results. GEO, generative engine optimization, is the work of making a page more visible inside answers that an AI system writes from several sources. Researchers at Princeton University and IIT Delhi, with two co-authors listed as independent in the KDD version, introduced the term in a paper first posted in November 2023 and accepted to KDD 2024. AEO, answer engine optimization, names the same goal from the answer side: being the source behind the answer an engine or assistant gives, rather than one link in a list.
Google treats the two newer terms as one job. Its guide to optimizing for generative AI features, last updated July 10, 2026, says AEO and GEO both describe work focused on visibility in AI search experiences. The same guide says that, from Google Search's point of view, that work is still SEO. Microsoft Advertising's October 2025 guidance on AI search answers, written by a Bing product manager, opens by treating GEO, AIO and SEO as names for one aim. Bing does use the word GEO for its own tooling: it calls its AI Performance report an early step toward GEO tools in Bing Webmaster Tools.
| Term | What it aims at | Where the term comes from | First-party measure |
|---|---|---|---|
| SEO | A high position in a search engine's ranked results | Long-standing industry term; Google and Bing document their own practices | Search Console's Performance report; Bing Webmaster Tools |
| AEO | Being the source behind an engine's or assistant's direct answer | Industry term; this article found no founding paper for it | No report is labelled AEO; the AI reports below cover it |
| GEO | Being cited, and taking up more of the answer, inside AI-written answers | Aggarwal and others, arXiv, November 2023; KDD 2024 | Search Console's generative AI report; Bing's AI Performance report |
How AI answers are built from search results
Google builds its AI answers from its search index. Its AI optimization guide says AI Overviews and AI Mode are rooted in Google's core Search ranking and quality systems, and it names two techniques. Retrieval-augmented generation, which Google also calls grounding, uses Search ranking to fetch relevant pages, then writes an answer that links to them. Query fan-out sends several related searches at once to gather more results for one question.
Google sets a floor for inclusion. A page must be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode, according to Google's AI features documentation. The site must also be included under the Search generative AI setting in Search Console. Google's help page for the control says Include is the default, though a property with a parent property inherits the parent's setting by default.
OpenAI and Perplexity mostly tie a site's place in their answers to their search crawlers. OpenAI's crawler documentation says OAI-SearchBot surfaces sites in ChatGPT's search features, and that sites which opt out of it are not shown in ChatGPT search answers, though they can still appear as navigational links. Perplexity's crawler documentation says PerplexityBot surfaces and links websites in Perplexity's search results and is not used to crawl content for AI foundation models. Its separate Perplexity-User fetcher, which acts on a user's question, can still visit and link a page and generally ignores robots.txt. Bing's June 16, 2026 post defines grounding as the source material and web evidence an AI answer uses to support and cite itself.
The consequence is simple. A page that an engine cannot fetch, or does not retrieve for the question, cannot be cited, however well it is written.
Does GEO replace SEO?
GEO does not replace SEO, according to both Google and a 2025 benchmark of GEO methods across six domains. Google's guide says SEO best practices stay relevant because its generative AI features are rooted in its core ranking and quality systems. The authors of C-SEO Bench reached the same conclusion by experiment. They write that conversational search optimization should complement traditional SEO rather than replace it, because a document's position in what the model reads outweighed every content rewrite they tested.
GEO does add something a ranked list never had to measure. A results page has positions, while an AI answer has citations, and several sources share one answer. Bing's AI Performance report therefore counts citations and cited pages rather than positions. Its June 2026 update added citation share: the percentage of all citations shown for a grounding query that went to your site.
Google also warns against a separate GEO playbook. Its guide lists tactics site owners can ignore for Google Search, including breaking content into small chunks for AI, rewriting pages just for AI systems and seeking inauthentic mentions.
What the research says actually moves citations
Two peer-reviewed studies sit behind most GEO advice, and their results differ mainly because they measured different things. The table sets them side by side from each paper's own text.
| Detail | GEO (Aggarwal and others) | C-SEO Bench (Puerto and others) |
|---|---|---|
| Published | arXiv, November 16, 2023; accepted to KDD 2024 | arXiv, June 6, 2025; accepted to NeurIPS 2025 Datasets and Benchmarks |
| Questions | 10,000 in GEO-bench (80% informational); results reported on its 1,000-question test split | More than 1,900 queries and more than 16,000 documents, across six domains in two tasks (product recommendation and question answering) |
| Answer engine | GPT-3.5 Turbo writing from the top five Google results; a 200-question check on Perplexity with the sources uploaded as files | GPT-4o mini, Claude 3.5 Haiku, o3 and o4-mini writing from a fixed set of documents |
| What it measured | How many words of the answer drew on a source, weighted by where its citation appeared | How early in the answer a source was cited |
| Headline result | Citations, quotations and statistics raised visibility 30 to 40%; keyword stuffing gave little or no gain | Significant gains in only 3 of 54 cases in the authors' main GPT-4o mini results, all in product recommendation, and none in question answering on any model; placing a document first in the model's input beat every rewrite |
The GEO paper's best methods added citations, quotations or statistics to a page. The paper reports a relative gain of 30 to 40% for those three on its main visibility measure. Keyword stuffing, which the paper treats as a classic SEO method, gave little or no improvement, and on Perplexity it scored about 10% below the unchanged page on the main measure. The gains went mostly to lower-ranked pages: when every source in the answer was rewritten with citations, the source ranked fifth in Google gained 115.1% in visibility, while the top-ranked source's fell by 30.3% on average.
Three details limit how far the GEO result carries. The main engine was a simulation, GPT-3.5 Turbo answering from five fetched pages. The Perplexity check covered 200 questions, and because Perplexity did not let users specify source URLs at the time, the authors uploaded the source text as files, so Perplexity's own retrieval was not part of the test. The best methods there gained 22% on the main measure and 37% on a second measure scored by a language model. And the abstract's figure, up to 40%, is a maximum, not a typical gain.
C-SEO Bench tested the same kinds of rewrite against citation order, on four models from OpenAI and Anthropic. Its authors report significant gains in only 3 of 54 cases in their main GPT-4o mini results, all in product recommendation, and no method helped question answering on any of the four models. Many rewrites made rankings worse: the statistics method lowered rankings in 19 of 24 settings. Gains also shrank as more competing documents adopted the same method, which the authors call a congested, zero-sum game.
The C-SEO Bench authors say the two papers do not contradict each other, because a page can take up more words in an answer without being cited earlier. This article's reading of the two is that being retrieved and ranked near the top matters more than how a page is phrased, and that phrasing tricks lose value as competitors copy them.
How to measure SEO, AEO and GEO
SEO, AEO and GEO each have a first-party report from Google or Microsoft, and neither company's report names ChatGPT or Perplexity among the surfaces it covers. Google's guide points to Search Console for AI visibility and warns that no third-party tool has access to Google's internal ranking or AI systems.
| Report | Covers | Measures | Breaks down by | Leaves out |
|---|---|---|---|---|
| Search Console Performance report | Google Search results | Clicks, impressions, click-through rate, average position | Queries, pages, countries, devices, search appearance, dates | AI feature traffic as its own line; Google includes it in these totals |
| Search Console generative AI performance report | AI Overviews and AI Mode; a separate report covers Discover | Impressions | Pages, countries, devices, dates | Clicks and queries: Google's help page lists neither |
| Bing Webmaster Tools AI Performance | Microsoft Copilot, AI summaries in Bing, select partner integrations | Citations, and cited pages per day on average | Grounding queries (a sample), cited URLs, a timeline; rolling out in preview from June 16, 2026: intents, topics, citation share and period comparison | Placement within an answer, competitor domains and traffic share |
| ChatGPT and Perplexity | ChatGPT search; Perplexity answers | No site-owner citation report is described in either company's crawler documentation | Not applicable | Citations: check by hand |
SEO maps to the Performance report. GEO and AEO in Google map to the generative AI report, which Google launched on June 3, 2026 for a subset of sites and rolled out to every site worldwide by August 31, 2026. An impression there is a link to your site shown in an AI feature, and the chart counts two results from the same site in one AI feature as a single impression. In Microsoft's products, GEO maps to AI Performance.
For ChatGPT and Perplexity, a fixed log of questions checked by hand is the fallback. This is this article's method, not any engine's, and it samples answers rather than measuring them:
- Write 20 questions a buyer asks before choosing a product like yours, in their words rather than your keywords.
- Ask each question in ChatGPT and Perplexity, in the same account state every time, and note that state in the log.
- Record the date, the engine, whether your brand is named, where it sits among the brands named, and every cited URL.
- Repeat monthly with the same questions. Bing's own team says citation patterns shift with models, freshness and demand, so read a trend across months, not one change in one run.
Where a small team should start
A small team should start with SEO, because retrieval decides whether any AI answer can use a page. Six steps cover the ground, in order:
- Let search crawlers in. Allow Googlebot and Bingbot, and allow OAI-SearchBot and PerplexityBot if you want to appear in ChatGPT search and Perplexity, per each company's crawler documentation.
- Keep pages eligible. Keep them indexable and eligible for a snippet, and leave Search Console's Search generative AI setting on Include. Excluding a site removes its links and content from Search generative AI features (AI Overviews, AI Mode and generative AI features in Discover), and Google says it then gets no traffic or impressions from them.
- Answer the questions buyers ask, one topic per page. Google's guide says creating separate content for every query variation or fan-out query, mainly to manipulate rankings or AI answers, violates its spam policies.
- Publish something nobody else has. Google's guide says unique, useful content, such as a first-hand review rather than a summary of other pages, will likely shape presence in generative AI search in the long run more than anything else it suggests.
- Cite sources and use real figures for readers, not as a lever. The GEO paper's lab gains from citations and statistics did not hold in C-SEO Bench, where the statistics method often lowered rankings.
- Verify your site in Search Console and Bing Webmaster Tools, and read both AI reports monthly. Bing's Compare view overlays the previous period on the current one, which makes a monthly read quicker.
The engine-specific steps are in our guides to earning citations in ChatGPT and getting cited in Perplexity.
Where overads fits
overads Signals is built for the GEO side of this work, on two engines. Signals is built to put tracked questions to Gemini and Perplexity on a schedule and read each answer: whether the brand is named, where it lands, how it is framed and which sites the answer drew on. Its AI-answer scan is switched off on the live product, so it returns no results for customers yet. Signals does not cover ChatGPT, Microsoft Copilot or Google AI Overviews.
overads does not replace the reports above. Google's guide says no third-party tool can see inside its ranking or AI systems, and that includes overads, so Search Console and Bing Webmaster Tools remain the record for Google's and Microsoft's surfaces. A team that needs ChatGPT or Copilot covered now needs a dedicated AI visibility tool, and our comparison of overads with Peec AI sets out what one such tool covers that Signals does not.
overads Publish posts to X, LinkedIn, Bluesky, Mastodon and Pinterest, which keeps a brand's own profiles current. Posts on your own accounts are owned content, though. They are not the discussion in blogs, videos and forums that Google's guide says its AI features can show.
Frequently asked questions
Does an llms.txt file help with GEO?
Not in Google Search. Google's AI optimization guide says Google Search ignores llms.txt and similar AI text files, so keeping one will neither harm nor help visibility there. Google adds that maintaining one for other services that read such files is fine.
Do I need special schema markup to appear in AI answers?
Not for Google. Its AI optimization guide says structured data is not required for generative AI search and there is no special schema.org markup to add, though structured data still helps a page qualify for rich results.
Is AIO the same thing as GEO?
It is another label for the same aim. Microsoft's October 2025 guidance on AI search answers opens by naming GEO, AIO and SEO side by side and says visibility is the common goal whichever name is used.
Sources
- GEO: Generative Engine Optimization (Aggarwal et al., KDD 2024). arXiv, Nov 16, 2023. Accessed Sep 27, 2026.
- C-SEO Bench: Does Conversational SEO Work? (Puerto et al., NeurIPS 2025 Datasets and Benchmarks). arXiv, Jun 6, 2025. Accessed Sep 27, 2026.
- Optimizing your website for generative AI features on Google Search. Google Search Central. Accessed Sep 27, 2026.
- AI features and your website. Google Search Central. Accessed Sep 27, 2026.
- Introducing Search Generative AI performance reports in Search Console. Google Search Central Blog, Jun 3, 2026. Accessed Sep 27, 2026.
- Generative AI performance report (Search). Search Console Help. Accessed Sep 27, 2026.
- Performance report (Search results): Overview and basic setup. Search Console Help. Accessed Sep 27, 2026.
- Search generative AI control. Search Console Help. Accessed Sep 27, 2026.
- Introducing AI Performance in Bing Webmaster Tools Public Preview. Microsoft Bing Blog, Feb 10, 2026. Accessed Sep 27, 2026.
- New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare. Microsoft Bing Blog, Jun 16, 2026. Accessed Sep 27, 2026.
- Optimizing Your Content for Inclusion in AI Search Answers. Microsoft Advertising, Oct 8, 2025. Accessed Sep 27, 2026.
- Overview of OpenAI crawlers. OpenAI. Accessed Sep 27, 2026.
- Perplexity crawlers. Perplexity. Accessed Sep 27, 2026.
- Signals. overads. Accessed Sep 27, 2026.
- overads vs Peec AI. overads. Accessed Sep 27, 2026.
- Publish. overads. Accessed Sep 27, 2026.
Drafted with AI assistance, then checked against primary sources and the product itself by the overads team.
Keep reading
How to rank in ChatGPT: what actually gets a brand cited
ChatGPT answers a buying question with a short list of names and a handful of cited pages. Here is what we have seen decide whether a brand makes that list, and a routine for getting there without gaming anything.
6 min readPerplexity SEO: how to get cited in Perplexity answers
Perplexity shows its sources on every answer, which makes it the easiest AI engine to study and the hardest to fake. Here is what we have seen get a page cited, and how to tell whether your share of those citations is moving.
6 min readGPT-6 Sol and Luna for marketers: what they change
OpenAI's two new GPT-6 models cost half or less of GPT-5.6's promotional API prices. Here is what that means for people who run ads, posts and reports, which model fits which job, and how to use them with the channels you already publish to.
7 min readSee how Gemini and Perplexity describe you
Signals is built to ask Gemini and Perplexity the questions your buyers ask and show whether, where and how you are cited. Its scan is switched off for now; sign up and we will tell you the day it is on.
Start free
