AI SEO has two meanings. The first is SEO for AI search. It means getting your brand named and your pages cited when ChatGPT, Gemini, Perplexity or Google AI Overviews answer a buyer's question. The second is using AI tools to do SEO work, such as research, drafts and checks.
Most of this guide is about the first meaning, where a brand can go missing without noticing. SEO (search engine optimisation) is the work of getting your pages found in search results. A Google AI Overview is the AI-written summary above the links on some Google searches.
What is AI SEO?
AI SEO is the work of making your brand easy for AI search tools to find, name and cite. It is also called AI search optimization, SEO for AI or AI engine optimization. A win is a mention inside the answer, backed by a citation: a link to a page that supports you.
Google does not treat AI search as a separate kind of work. Its guide for site owners says that, from Google Search's point of view, optimising for AI search is still SEO (Google Search Central, checked 7 Oct 2026). The newer names still help. They point at a result classic SEO reports do not show: whether the answer names you.
How is AI SEO different from classic SEO, GEO, AEO and LLM SEO?
Classic SEO earns your page a place in a list of links. AI SEO, GEO, AEO and LLM SEO are four names for one newer goal. That goal is getting your brand named and your pages cited inside an AI-written answer.
| Term | Stands for | What you win | About the name | Read more |
|---|---|---|---|---|
| Classic SEO | Search engine optimisation | A place in the list of links, and the click | The original name for search work | GEO vs SEO |
| AI SEO | SEO for AI search | Your brand named and cited inside AI answers | The broadest label, also used for SEO done with AI tools | This guide |
| GEO | Generative engine optimisation | Same as AI SEO | From a research paper first posted on 16 November 2023 (arXiv, checked 7 Oct 2026) | What is GEO |
| AEO | Answer engine optimisation | Same as AI SEO | Treats AI tools as answer engines | What is AEO |
| LLM SEO | SEO for large language models | Same as AI SEO | Named after the AI that writes the answers | LLM SEO |
What changes and what stays the same
A win is now a mention, which may come with no click. Results are judged over many answers, because answers change between runs. LLM SEO covers these changes in full.
The base stays the same: search engines must be able to reach your pages and store them, and the pages must be useful. GEO vs SEO and AEO vs SEO cover what carries over.
How do AI engines pick the sources they cite?
AI engines cite pages that their search systems can find and are allowed to read. Google, OpenAI and Perplexity document how that works and what keeps a page out. We found no page from any of them that says how a brand gets picked.
Google AI Overviews
Google says its AI features are built on its core Search ranking systems. Search first finds relevant pages in its index, the database of pages Google keeps. The AI then writes a response from them and links to pages that support it. The model may also run several related searches at once, called query fan-out (Google Search Central, checked 7 Oct 2026).
To be a supporting link, a page must be indexed, meaning stored in that database. It must also be eligible to show with a snippet, the preview text under a result. Google lists no additional technical requirements (Google Search Central, checked 7 Oct 2026). The site must be included under the Search generative AI control in Search Console, Google's tool for site owners. Included is the default (Search Console Help, checked 7 Oct 2026). Guide: how to rank in Google AI Overviews.
ChatGPT search
OpenAI says any public website can appear in ChatGPT search (OpenAI Help Center, checked 7 Oct 2026). Its search crawler, a bot that fetches web pages, is called OAI-SearchBot. Sites that opt out of it are not shown in ChatGPT search answers, though they can still appear as navigational links. A separate crawler, GPTBot, collects content that may be used for training, and OpenAI says the two settings are independent (OpenAI crawler documentation, checked 7 Oct 2026). Guide: how to get ChatGPT to recommend your brand.
Perplexity
Perplexity says PerplexityBot is designed to show and link websites in its search results. It is not used to crawl content for AI foundation models, the large models that AI tools are built on. Perplexity recommends allowing it in robots.txt, the text file that tells each bot what it may read. A second bot, Perplexity-User, visits a page when a user's question needs it and generally ignores robots.txt rules (Perplexity crawler documentation, checked 7 Oct 2026). Guide: how to rank in Perplexity.
Gemini
For the Gemini app, Google documents a robots.txt setting called Google-Extended. It decides whether your content may be used to train future Gemini models and for grounding in Gemini Apps. Grounding means giving the model content from the Google Search index as it answers. Google-Extended does not affect inclusion or ranking in Google Search (Google crawler documentation, checked 7 Oct 2026). Guide: how to rank in Gemini.
How do you start AI SEO? A five-step plan for a consumer brand in India
Start with access, then questions, then a first measurement, then fixes, then a second measurement. The first three need only a browser and a spreadsheet. Step 1 follows what the engines document. The off-site work in Step 4 is our own recommendation.
Step 1: Check that AI search crawlers can reach your site
Open yourdomain.com/robots.txt and search for three names: Googlebot, OAI-SearchBot and PerplexityBot. They are the search crawlers for Google, ChatGPT and Perplexity. A Disallow: / line under a name closes the whole site to that crawler, so have it removed. If a name is not in the file, the group that starts with User-agent: * applies to that bot (RFC 9309, checked 7 Oct 2026). Search for Google-Extended too. The Gemini section above explains what a Disallow rule under it does.
Google says robots.txt rules for Googlebot control its AI features too. It also asks that crawling is allowed by any CDN, the service that delivers your pages, and by your hosting setup (Google Search Central, checked 7 Oct 2026). A firewall or CDN can refuse a bot that robots.txt allows. To check, have your developer search the server logs for each bot name. A 403 status, the "access denied" reply, beside a name means the bot was refused. The AI crawlers list explains each bot, and Depra AI's AI crawler checker reads your robots.txt for 41 AI crawlers.
Step 2: Write down the questions buyers ask
List 20 to 40 buying questions for your category, written the way a shopper types them. One example: 'best sunscreen for oily skin under ₹500'. Take them from support chats, reviews and your own site search.
Leave your brand name out, because a question that names you will almost always return you. If your buyers write in Hinglish, which is Hindi and English typed in English letters, add those versions. Then keep the list fixed.
Step 3: See who gets named
Ask each question in ChatGPT with web search on, Gemini, Perplexity and Google, in a private window. Record whether you were named, your place in the list, the rivals named and the websites cited. On Google, note whether an AI Overview appeared at all. Google says Overviews often do not trigger (Google Search Central, checked 7 Oct 2026).
Ask each question at least twice, on different days. Depra AI's study collected 480 answers in India on 14 August 2026. It compared two runs of the same question in the same language. On average they shared 54.7% of their named brands on ChatGPT, 51.2% on Gemini and 68.3% on Perplexity (Hinglish vs English AI shopping study). The share of answers that name you is your AI visibility.
Step 4: Fix your pages and the third-party sources
On your own site, make each product and category page answer the buying question near the top. Add facts a shopper can check: price, size, ingredients and who it suits. Google advises a point of view of your own, such as a first-hand review, over a summary of what others say. It also says Merchant Center feeds, the product data a store sends to Google, can help products show in AI responses (Google Search Central, checked 7 Oct 2026).
Off your site, open the pages the answers cited. In Depra AI's study, youtube.com had 534 of the 4,640 citations, reddit.com 191 and nykaa.com 176 (Hinglish vs English AI shopping study). Note the cited videos, threads and marketplace pages that name your rivals and not you. Send list authors and video reviewers your product's facts and a sample. Answer open forum threads from a named brand account, within each forum's rules. Create or claim your marketplace listing. Depra AI's citation gap analysis lists the sites cited in answers that name a competitor while your brand is missing. The pricing page lists it on Growth and above. See also marketplace visibility in AI answers.
Step 5: Measure again
Run the same questions on the same engines a few weeks later and compare totals, not single answers. Small moves can come from run-to-run changes alone. How to measure AI visibility explains how many answers you need.
Can you use AI tools to do SEO work?
Yes, for research, structure and first drafts, as long as a person checks every fact before it goes live. This second meaning of AI SEO is also called AI-powered SEO or SEO with AI. For software, see the best AI SEO tools and SEO automation.
Google's guidance says generative AI, the kind that writes text, can help with researching a topic and structuring original content. Generating many pages without adding value may violate its spam policy on scaled content abuse. All AI-generated content should be fact-checked by a person before publishing (Google Search Central, checked 7 Oct 2026). In practice:
- Good uses: grouping buyer questions into topics, outlining a page, drafting titles and meta descriptions for a person to edit, and summarising customer reviews.
- Poor uses: writing prices, claims or statistics with no source, and publishing pages in bulk with no human edit.
- What a chatbot cannot tell you: how visible your brand is. One chat cannot show what thousands of buyers were told.
How do you measure AI SEO results?
Measure two things side by side: what Google and ChatGPT report about your pages, and what the AI answers say about your brand.
What Google and ChatGPT report
Search Console has a Generative AI performance report. It shows how many times links to your site were shown in AI Overviews and AI Mode, Google's chat-style search mode. Data is grouped by page, country, date and device (Search Console Help, checked 7 Oct 2026). It counts links to your pages. The help page lists no count of brand mentions.
For ChatGPT, OpenAI says it adds utm_source=chatgpt.com to the links that send visitors to your site. Those visits show under that source in an analytics tool (OpenAI Help Center, checked 7 Oct 2026).
What the answers say about your brand
This part you collect yourself or with a tracker. Count AI visibility, the share of answers that name you. Add AI share of voice, your share of all tracked brand mentions. Then note your position in the list and the websites cited.
Measuring with Depra AI
Depra AI's AI visibility tracking asks your buying questions to ChatGPT with web search, Gemini and Google AI Overviews every day. Perplexity is checked weekly. It asks from inside the country you choose, in English and Hinglish. Every answer is stored with the brands named and the sites cited. Plans start at ₹1,999 a month plus GST.
Depra AI is software, not an agency. It does not write content, edit your website or report traffic and sales from AI answers. Google AI Mode, Claude and Microsoft Copilot are not tracked.
What are the common AI SEO mistakes?
Most mistakes come from treating AI SEO as a set of tricks, or judging it on too little evidence.
- Blocking the wrong bot. A rule meant to stop AI training can also shut out search. Block GPTBot if you wish, and keep OAI-SearchBot, PerplexityBot and Googlebot allowed.
- Judging from one chat. One answer is a sample of one. Count across many.
- Dropping classic SEO. Google says a page must be indexed to be a supporting link in an AI Overview (Google Search Central, checked 7 Oct 2026). Getting pages indexed is classic SEO work.
- Chasing special files. Google says Google Search ignores llms.txt, a proposed text file that points AI tools to a site's key pages. It also says content need not be split into small pieces or rewritten for AI (Google Search Central, checked 7 Oct 2026). See the llms.txt guide.
- Checking English only. In Depra AI's study, Perplexity's brand lists overlapped 23.4 points less across English and Hinglish than across two runs in one language (Hinglish vs English AI shopping study).
- Expecting a guarantee. No one can promise a place in an AI answer. Google says even indexing is not guaranteed (Google Search Central, checked 7 Oct 2026).