AI SEO is the practice of optimising content so that AI answer engines such as ChatGPT, Gemini, Google AI Overviews and AI Mode find, trust and cite it. The same term also describes using artificial intelligence to do SEO work. A complete AI SEO strategy covers both.
Search has split in two. People still type queries into Google, but they increasingly ask AI systems that write an answer and cite a handful of sources. Those systems lean on the same foundations as classic search, and Google says so plainly[1]. They also behave differently in ways that change what you optimise and how you measure it. This guide separates what is genuinely new from what is SEO with a new name, and it quotes Google directly where Google has said something.
What AI SEO means (both meanings)
Search for “AI SEO” and you will find two different conversations, often in the same article. It helps to separate them before doing anything:
Why AI SEO matters now
The audience for AI answers is now measured in billions. These are the latest figures the companies themselves have published:
Sources: AI Overviews and AI Mode[5], Gemini app[6], ChatGPT[7]. Each company defines a “user” its own way, so the figures are not directly comparable.
Traffic tells a more modest story, and it is worth being honest about it. Across sites that use Ahrefs’ free analytics, AI assistants sent a small but steadily growing share of all visits:
| Period | AI share of visits |
|---|---|
| Mar 2025 | 0.1% |
| Jun 2025 | 0.25% |
| Mar 2026 | 0.28% |
| Aug 2026 | 0.41% |
Under half a percent of visits sounds small, and for most sites it is still small next to Google. But it undercounts the effect: many people read an AI answer, take the recommendation and never click. Whether your brand is named in the answer matters more than the click. Among the AI referrals that do happen, one assistant dominates, although its lead is shrinking:
| ChatGPT | 81.4% |
|---|---|
| Gemini | 11.6% |
| Perplexity | 4.6% |
| Claude | 2.3% |
Professionals feel the shift. In SEOFOMO’s August 2026 survey of 171 search professionals, 98% said clients or managers had asked them about AI search visibility, and 69% now have a dedicated budget for it[16].
How AI search picks the answers it gives
Google describes its AI features as retrieval-augmented generation, “also known as grounding”: its core ranking systems retrieve pages from the index, a model writes a response, and links to supporting pages are shown[1]. ChatGPT works in a similar way, rewriting your question into one or more targeted searches that it sends to search providers[8]. Step through it:
Three consequences follow, and they are what separate AI SEO from classic SEO:
- You compete on sub-questions. A fan-out creates many smaller searches. Your page can win one of them without ranking for the question the person typed.
- Several sources share one answer. Being one of five cited sources is the goal, and a lower-ranked page can be cited alongside the top result. In the GEO study, pages at the fifth retrieved position gained the most from optimisation[4].
- The sources move. SISTRIX found that Google AI Mode replaces 56% of its cited domains every week and ChatGPT Search 74%, while AI Overviews change only about 5%[14]. A single check tells you little.
What Google says actually works
In July 2026 Google published an AI optimization guide that is refreshingly direct. Its core message: “Is SEO still relevant for generative AI search? In short, yes!”, because its generative AI features “are rooted in our core Search ranking and quality systems”[1]. It also warns that “many suggested ‘hacks’ aren’t effective or supported by how Google Search actually works” and tells site owners to “prioritize effective SEO strategies over ‘AEO/GEO hacks’”[1]. Several popular tactics fall away as a result:
What Google does recommend is familiar: allow crawling, make content findable through internal links, give a good page experience, put important content in text, use good images and video, keep structured data consistent with visible text, and keep your Business Profile current[2]. To be eligible as a supporting link at all, a page must be “indexed and eligible to be shown in Google Search with a snippet”[2].
SEO for AI: how to get cited by AI answers
With the myths set aside, these are the practices with evidence behind them. They apply across AI Overviews, AI Mode, ChatGPT, Gemini, Claude and Perplexity.
1. Let the right crawlers in
Each AI engine uses its own crawler, and some run separate crawlers for training and for search. OpenAI says sites that block OAI-SearchBot will not appear in ChatGPT search answers, while GPTBot only controls training[8]. Google’s AI features use Googlebot. Check your file with the AI Crawler Checker, and fix it with the robots.txt generator.
2. Answer the sub-questions, answer-first
Map the questions a buyer would ask around your topic: comparisons, prices, alternatives, who it suits, how it works. Give each a clearly headed section that opens with a direct, self-contained answer. That is the passage a model can lift, and it is also what makes a page easy to scan. Our answer engine optimization guide covers the structure in detail.
3. Add evidence: statistics, quotes, sources
The best-known controlled study of GEO tactics, published at KDD 2024, found that adding quotations, statistics and citations to sources raised a page’s visibility in generated answers by roughly 30–40% on its main measure, while keyword stuffing gave “little to no improvement”[4]. The study measured pages already retrieved as one of five sources, so it shows how to be more prominent once found, not how to be found. The GEO vs SEO guide breaks down each tactic’s result.
4. Be present where AI engines look
AI answers draw heavily on third-party sources: reviews, forums, comparison articles, video and reference pages. Google itself says its AI features “can show what’s being said about products and services across the web, including in blogs, videos, and forum discussions”[1]. In AI Mode, SISTRIX found YouTube and video domains hold a place in the stable core of an answer far more often than news sites do[14]. Find which sources AI cites for your category, and earn a presence in them.
5. Keep facts consistent everywhere
AI engines build a picture of your brand from many pages. Consistent names, descriptions, prices and category terms on your site, profiles and listings make it easier for them to describe you correctly, and inconsistent ones invite errors. The E-E-A-T post, E-E-A-T when the reader is a machine, explains which trust signals a model can actually parse.
LLM optimization beyond Google
LLM optimization is the same work pointed at every large language model, not just Google’s. Two things make it different from optimising for one search engine.
Each engine has its own index and its own habits. A page can be cited by Perplexity and ignored by ChatGPT. SISTRIX measured the domain overlap between AI Mode and ChatGPT for the same prompts at just 0.125 on a 0-to-1 scale[14]. Visibility has to be checked per engine.
Recommendations are inconsistent by design. In a SparkToro and Gumshoe study of 2,961 runs of the same prompts, there was less than a 1 in 100 chance that ChatGPT or Google’s AI would return the same list of brands twice. The study concluded that visibility across many prompts, run many times, is a reasonable metric, while a single “ranking position in AI” is not[15]. For the engine-specific playbooks, see the ChatGPT SEO, Google AI Mode and Claude SEO guides.
AI for SEO: which tasks AI tools actually help with
The other half of AI SEO is using AI to do the work. Google’s position since 2023 is that it rewards helpful content however it is produced, and that automation used primarily to manipulate rankings breaks its spam policies[3]. That makes AI a legitimate tool for most SEO tasks, as long as a person owns the result. This is where it helps, and where it needs a human:
| AI helps a lot | Needs a human | Main risk | |
|---|---|---|---|
| Keyword research and clustering | ✓Yes | ~ | Invented volumes if the tool has no real data |
| Content briefs | ✓Yes | ~ | Briefs that copy what already ranks |
| First drafts | Good first draft | ✓Yes | Generic text with no original insight |
| Meta titles and descriptions | ✓Yes | ~ | Over-length or duplicated tags |
| Internal linking suggestions | ✓Yes | ~ | Links to pages that no longer exist |
| Schema markup | ✓Yes | ~ | Markup that does not match the visible page |
| Technical audits | ~ | ✓Yes | Plausible fixes for problems that are not there |
| Original research and expertise | ✕No | ✓Yes | Fabricated statistics and quotes |
For tool recommendations by job, with prices read from each vendor’s own page, see our comparison of the best AI SEO tools. Our free tools cover several of these tasks without an account: the schema markup generator, the llms.txt generator and the AI SEO checker.
An AI SEO strategy, step by step
Put the two halves together and you get a strategy with a clear order. Work through it top to bottom; your ticks are saved in this browser.
How to measure AI SEO
Measurement is where most AI SEO programmes go wrong, because the habits of rank tracking do not transfer. An AI answer has no fixed position, its sources change weekly, and many answers are read without a click. Three signals are worth tracking together:
- Visibility. The share of your tracked questions in which each engine names you, measured repeatedly. This is the metric the SparkToro study endorses and the one AI visibility tracking reports.
- Durability. Whether you hold a stable place in the answer or rotate in and out. Our Core vs Carousel view separates the two.
- Traffic. AI referrals in your analytics and AI impressions in Search Console. These are useful but undercount AI’s effect. The guide to measuring AI traffic in GA4 shows how to set it up.
Where do you stand? A two-minute check
Answer honestly. The result points you at the part of the strategy to start with.
Questions people ask
What is AI SEO?
AI SEO has two meanings. The first is optimising content so AI answer engines such as ChatGPT, Google AI Overviews, AI Mode, Gemini, Claude and Perplexity find, trust and cite it. The second is using AI tools to do SEO work faster, such as keyword research, content briefs, drafting and audits. Most strategies need both.
Is SEO still relevant with AI search?
Yes. Google's own AI optimization guide answers this directly: "Is SEO still relevant for generative AI search? In short, yes!", because its AI features are rooted in its core Search ranking and quality systems. Pages that are crawlable, indexable, useful and clearly structured are the ones AI features can draw on.
What is the difference between AI SEO and traditional SEO?
Traditional SEO aims for a position in a list of links. AI SEO aims to be one of the sources an AI system retrieves and cites when it writes an answer, and to be mentioned by name. The foundations overlap heavily, but AI answers fan a question out into many searches, cite several sources, and change their sources far more often, so the tactics and the measurement differ.
What is LLM optimization?
LLM optimization, also called LLMO or LLM SEO, is the practice of making a brand and its content more likely to be surfaced accurately by large language models, both in answers grounded in live search and in what models absorb in training. In practice it overlaps almost entirely with GEO and AEO. No consensus definition separates the terms.
Does Google penalise AI-generated content?
Not for being AI-generated. Google's guidance since February 2023 is that it rewards helpful, reliable content however it is produced, and that using automation to generate content primarily to manipulate rankings violates its spam policies. The quality and purpose of the page matter, not the tool used to write it.
Do I need llms.txt for AI SEO?
Not for Google. Google's AI optimization guide says Google Search itself does not use llms.txt files. Other AI companies have not committed to reading it either. It costs little to add, but it is not a ranking lever, and anyone selling it as one is overstating.
Does structured data help with AI search?
It helps machines understand a page, which is useful, but Google says structured data isn't required for generative AI search and there's no special schema.org markup you need to add. Keep any markup accurate and consistent with the visible page.
How do I optimize for ChatGPT?
Make sure OpenAI's search crawler, OAI-SearchBot, is not blocked in robots.txt, because OpenAI says sites that block it are not shown in ChatGPT search answers. Then work on being present in the sources ChatGPT searches and cites for your category, and on pages that answer specific questions directly. Our ChatGPT SEO guide covers it in detail.
Can AI do SEO for me?
It can do much of the labour: clustering keywords, drafting briefs and first drafts, writing meta descriptions, spotting technical issues and summarising data. It cannot decide what is worth saying, supply original experience or data, or judge whether a page is genuinely better than what already ranks. Treat AI output as a draft that a knowledgeable person edits.
How do I measure AI SEO?
Track whether AI engines mention and cite you for the questions that matter, repeated over weeks, because single checks are unreliable. Alongside that, watch AI referral traffic in your analytics and Google's AI impressions in Search Console. Traffic alone understates AI visibility, since many answers are read without a click.
What are the best AI SEO tools?
It depends on the job: content optimisation tools such as Surfer, Clearscope and Frase; suites with AI features such as Semrush and Ahrefs; and AI visibility trackers that monitor ChatGPT, Gemini and AI Overviews. Our comparison of the best AI SEO tools lists them with prices read from each vendor's own page.
What is an AI SEO strategy?
A plan covering both halves: technical access (crawlers allowed, pages indexable), content that answers the questions your buyers ask AI, presence in the third-party sources AI engines cite, AI tools to speed up production, and measurement of AI visibility over time. The strategy checklist on this page walks through each part.
Sources
Google and OpenAI documentation and company announcements are primary sources. The GEO paper is peer-reviewed research. Ahrefs, BrightEdge, SISTRIX, SparkToro and SEOFOMO are independent studies with their own methods, dated below.