If you run a business website, you’ve probably been told that search is changing — that people now ask ChatGPT, Copilot, Gemini or Google’s AI features instead of scrolling through links, and that you need GEO (Generative Engine Optimization) instead of, or as well as, SEO.
Some of that is true. A lot of what’s being sold around it isn’t. This article separates what search providers actually say from what’s speculative, so you can make sensible decisions without paying for promises nobody can keep.
Short version: the foundations of good SEO — crawlable pages, clear and accurate content, sound technical setup, and genuine usefulness — remain the foundations of being found and cited by AI search features. There’s no secret file, special markup or paid shortcut that guarantees placement in AI answers. And anyone who promises otherwise is promising something they don’t control.
A few definitions
Search engine optimization (SEO) is the practice of making your website easy for search engines to crawl, understand and present to people — and making it worth presenting. It covers technical foundations (can the pages be found and read?), content (does it answer people’s questions well?) and reputation (do other credible sources reference you?).
Crawling and indexing are the first two steps. A search engine’s crawler (Google’s is Googlebot; Bing’s is Bingbot) fetches pages and follows links. Indexing is storing and organizing what it found so it can be retrieved later. A page that isn’t crawled or indexed can’t appear in results — including AI-generated ones that rely on the search index.
Generative Engine Optimization (GEO) is a newer term for efforts to make content more likely to be used, cited or referenced in AI-generated answers. The term was popularized by a 2023 research paper, GEO: Generative Engine Optimization (Aggarwal et al.), which found that certain content characteristics — such as including citations, quotations and statistics from credible sources — were associated with greater visibility in the generative engines the researchers tested. It’s useful research, but it’s early, and results in experimental settings don’t automatically transfer to every commercial AI product.
Answer Engine Optimization (AEO) is an older, overlapping term focused on structuring content so it directly answers specific questions — the kind of content that appears in featured snippets, voice assistants and now AI answers. In practice, AEO and GEO advice overlap heavily with good SEO and good writing.
What Google actually says
Google is unusually direct on this point. Its Search Central documentation on AI features and your website states that SEO best practices remain relevant for AI Overviews and AI Mode, and that there are no additional requirements or special optimizations needed to appear in them.
The same page says:
- A page must be indexed and eligible to appear in Google Search with a snippet to be shown as a supporting link in AI Overviews or AI Mode. There are no additional technical requirements.
- You don’t need new machine-readable files, “AI text files” or special markup, and there’s no special schema.org structured data you need to add.
- Traffic from AI features is included in Search Console’s Performance report under the “Web” search type, rather than being reported separately.
- Meeting the requirements doesn’t guarantee that a page will be crawled, indexed or shown.
Google points site owners back to fundamentals: allowing crawling (including at the CDN or hosting layer), making content findable through internal links, providing a good page experience, keeping important content in text form, making sure structured data matches visible text, and creating helpful, reliable, people-first content.
What Microsoft Bing says — and now reports
Bing’s Webmaster Guidelines emphasize similar fundamentals: crawlable pages, clear content written for people, and avoiding manipulative techniques.
Bing has gone further than most on measurement. In February 2026 it introduced AI Performance in Bing Webmaster Tools as a public preview, showing when a site is cited in AI-generated answers across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. It reports citation counts, cited pages and sampled “grounding queries”. Bing is careful to note that these figures describe citation patterns and don’t indicate ranking or a page’s role within an individual answer.
That’s an important development: for at least one major provider, AI citation visibility is now something you can actually observe rather than guess at.
What other AI search providers say
AI products that search the web use their own crawlers, and some providers document them. OpenAI, for example, describes its crawlers separately:
- OAI-SearchBot is used to surface websites in ChatGPT’s search features. OpenAI says sites that block it won’t be shown in ChatGPT search answers (though they may still appear as navigational links).
- GPTBot is used to crawl content that may be used to train OpenAI’s foundation models. Blocking it is a training opt-out, and it’s controlled independently of search.
- ChatGPT-User visits pages when a person asks ChatGPT to, and OpenAI notes that robots.txt rules may not apply to these user-initiated requests.
The lesson generalizes: “AI bots” are not one thing. Search crawlers, training crawlers and user-initiated assistants serve different purposes. Blocking every AI-related crawler to protect your content may also remove you from AI search results you wanted to appear in. Make that choice deliberately, crawler by crawler.
Google offers a similar distinction: the Google-Extended control lets sites limit use of their content for some of Google’s AI training and grounding outside Search, while Google states it doesn’t affect inclusion in Google Search.
If your site uses Cloudflare, its bot-management and AI crawler settings can block categories of crawlers at the network level. That can be useful, but check the settings so you don’t block search crawlers you depend on.
What about llms.txt?
llms.txt is a proposed convention: a Markdown file at the root of a website that summarizes its content for large language models. It’s an interesting idea, and it costs little to publish.
But it’s a proposal, not a standard that the major search providers require. Google’s documentation explicitly says you don’t need new AI text files to appear in its AI features. If you publish an llms.txt, treat it as an optional experiment with no guaranteed effect.
What can realistically be measured
Be wary of anyone who reports precise “AI visibility scores” without explaining exactly how they’re produced. Here’s what you can measure with reasonable confidence today:
- Google Search Console — impressions, clicks and queries from Google Search, which includes AI Overviews and AI Mode traffic within the overall Web figures.
- Bing Webmaster Tools — traditional search performance, plus AI Performance citation data for Copilot and Bing AI summaries where available.
- Your own analytics — visits referred from AI products (some, though not all, pass referrer information), and, most importantly, enquiries and conversions.
- Manual citation checks — periodically asking AI assistants the questions your customers would ask, and recording whether and how your business is mentioned. Because AI answers are probabilistic and change over time, treat this as a qualitative sample, not a ranking report. Record the date, the exact question and the product used.
What you generally can’t measure reliably is your “rank” inside an AI answer. Answers vary between users, sessions and wording, and providers don’t publish placement data in the way traditional search rankings could be tracked.
What businesses should avoid believing
- “We can get you recommended by ChatGPT.” No outside party controls what an AI model says. Good content and sound technical foundations improve the odds of being found and cited; they don’t guarantee it.
- “You need special AI schema or files.” Google says there’s no special structured data or file required for its AI features. Accurate structured data that matches your visible content is still good practice — but as SEO hygiene, not a secret switch.
- “SEO is dead.” AI search features from Google and Bing are built on top of their search indexes. Being crawlable, indexable and useful matters at least as much as before.
- “More content is better.” Thin, repetitive, keyword-stuffed or mass-generated pages are what search providers’ guidelines warn against. A smaller number of genuinely useful pages is a better investment.
- “Block all AI bots to be safe.” That may also block AI search visibility. Decide what you’re actually trying to prevent — usually model training — and configure accordingly.
What actually helps
None of this is exotic. It’s the kind of work that makes a website better for people and easier for any system to understand:
- Make sure you can be crawled. Check robots.txt, CDN and firewall bot settings, and that important pages are reachable through normal links.
- Say clearly who you are and what you do. Use consistent business information across your site and other profiles. Explain services in plain text, not only in images or scripts.
- Answer real questions directly. Each important page should give a clear answer to its main question early, then support it with detail.
- Be specific and accurate. Original explanations, real examples, clear limitations and honest claims. Invented statistics and testimonials damage trust with people and systems alike.
- Cite credible sources where you make factual claims, and keep content up to date with visible review dates.
- Use structured data accurately — Organization, Article, Breadcrumb and similar types — describing only what is visibly true on the page.
- Measure what you can, using Search Console, Bing Webmaster Tools and your own enquiry data, plus documented manual checks.
The bottom line
Generative search is a real change in how people discover information, and it’s worth paying attention to. But the evidence so far points to an unglamorous conclusion: the businesses most likely to be found and cited by AI systems are the ones with clear, accurate, useful, technically sound websites — the same qualities that good SEO has always rewarded.
If someone offers you guaranteed placement in AI answers, ask them how they control the model. They can’t.
We built this website using the approach described here. If you’d like an honest, evidence-based look at your own site’s discoverability, get in touch.
This article is general information, not advice about your specific situation. We review our articles periodically and update them when the facts change.