AI Search Optimization: How to Get Your Business Cited in Google AI Overviews, ChatGPT and Perplexity

What is AI search, and why does it matter for your business?

AI search is any search experience where the first thing a person sees is a written answer put together by a language model, with links to the pages it drew on. Google shows these as AI Overviews and AI Mode. ChatGPT, Perplexity and Microsoft Copilot do the same thing inside a chat window. The person asks a full question, and the system replies in sentences instead of a list of ten blue links.

For a business, the change is simple to state. In a normal results page you compete for a position. In an AI answer you compete to be one of the few sources the answer quotes, names or links to. If a buyer in Pune asks an assistant which kind of agency can rebuild a slow Shopify store, the assistant will mention a handful of options at most. Being absent from that short list is the new version of being on page two.

The work of earning that place has a few names. Generative engine optimisation (GEO) and answer engine optimisation (AEO) are the common ones, and we use AI search optimisation as the plain-English label. Whatever you call it, it is not a replacement for SEO. It is SEO done with one extra question in mind: could a machine lift a correct, complete answer from this page and know who said it?

How do AI search engines decide which pages to cite?

Most AI search products work in two steps. First they retrieve pages from a search index. Then a language model reads those pages and writes an answer, attaching links to the passages it relied on. The model is not inventing its list of sources from memory. It is choosing among pages that a search system has already found, crawled and judged relevant.

Google describes one detail of this openly. For AI Overviews and AI Mode it may use a technique it calls query fan-out: the system breaks a question into several related searches, runs them, and combines what it finds. A question such as "how much does a website cost for a small clinic" may fan out into searches about website pricing, clinic website features and appointment booking. A page that answers one of those sub-questions clearly can be cited even if it does not rank first for the original wording.

The other assistants follow the same pattern with different indexes. Microsoft Copilot draws on the Bing index. ChatGPT search uses its own crawler together with third-party search providers. Perplexity runs its own crawler and index. This is why the first job is always the unglamorous one: make sure your pages can be crawled, are indexed, and say something specific.

• Retrieval comes first: a page that is blocked, not indexed or hidden behind scripts cannot be cited.
• Passages matter more than pages: assistants quote the paragraph that answers the question, so each section should stand on its own.
• Agreement across sources helps: when several independent sites say the same thing about you, the answer is safer for a model to repeat.

Do you need to do anything special for Google AI Overviews?

No. Google's own documentation is direct on this point: there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimisation is needed beyond what already makes a page eligible for Search. A page has to be indexed and eligible to be shown with a snippet. The same guidance on helpful, reliable, people-first content applies.

That is useful to know because a lot of advice in this area sells shortcuts. You do not need a special schema type, an "AI sitemap" or a text file written for language models to be considered. Google has said it does not use llms.txt. If you want less of your content used in AI features, the controls are the ones you already know: nosnippet, data-nosnippet, max-snippet and noindex.

One name causes confusion. Google-Extended is a robots.txt token that controls whether your content may be used to train and ground Google's Gemini models. Google states that it does not affect how a site is included or ranked in Google Search. Blocking it will not remove you from AI Overviews, and allowing it will not put you there.

Step 1: Can AI crawlers actually read your website?

Start with access, because everything else depends on it. Each AI search product reads the web with a named crawler, and each crawler obeys robots.txt. Open yourdomain.com/robots.txt and look for rules that block more than you intended. We often find a blanket block added years ago to stop scrapers, or a security plugin that challenges every visitor it does not recognise.

It helps to separate two kinds of crawler. Search crawlers fetch pages so they can be shown and linked in answers. Training crawlers collect text to train future models. OpenAI, for example, documents three: OAI-SearchBot for ChatGPT search results, GPTBot for model training, and ChatGPT-User for pages a person asks ChatGPT to open. You can allow search and decline training if that is your policy. Blocking the search crawler, though, removes you from that product's answers.

Then check the layer above robots.txt. A firewall or content delivery network can block bots that robots.txt allows. Ask whoever manages your hosting to confirm that verified search crawlers are not being challenged or rate-limited. Finally, make sure your important text is in the HTML the server sends. Content that only appears after heavy scripts run, or that lives inside images and PDFs, is easy for a crawler to miss. This is ordinary technical SEO, and it is the cheapest fix on this list.

# Allow search and AI search crawlers, keep utility paths private
User-agent: *
Allow: /
Disallow: /thank-you/

# Optional: decline model training while staying visible in ChatGPT search
User-agent: GPTBot
Disallow: /

Sitemap: https://www.example.com/sitemap.xml

Step 2: How should you write so an AI can quote you?

Write answer-first. Put the question in the heading, in the words a customer would use, and answer it in the first one or two sentences underneath. Detail, conditions and examples come after. A model looking for a passage to quote will find a complete answer at the top of the section, and so will a person skimming on a phone.

Make each section self-contained. A paragraph that begins "As mentioned above, it depends" is useless when lifted out of the page. Name the thing you are talking about again, give the answer, and state the main condition. Short paragraphs with one idea each are easier to quote accurately than long ones that wander.

Be specific where you honestly can. "Timelines vary" tells nobody anything. "A business website typically takes two to three months from discovery to launch" is a sentence an assistant can use, and it is the kind of statement that sets one source apart from twenty that hedge. Do not invent numbers to sound precise. A figure with no basis is worse than no figure, and it is exactly what a careful reader, human or machine, will check against other sources.

• Use question headings: "How long does SEO take to show results?" instead of "Timelines".
• Give a 40 to 60 word direct answer at the top of the page for the main question.
• Use lists for steps and options, and tables for comparisons.
• Define terms the first time you use them, in one plain sentence.
• Keep one topic per page so the page has a clear subject.

Step 3: How do you show experience and trust (E-E-A-T)?

Google's quality guidelines describe experience, expertise, authoritativeness and trust, usually shortened to E-E-A-T. It is not a score you can read anywhere. It is a description of what trustworthy pages have in common, and AI systems lean on the same signals because they need sources they can attribute an answer to.

Begin with identity. Say who is behind the website: the legal business name, a real address, a phone number, and an About page that explains what you do and for whom. Use the same name, address and phone number everywhere they appear. A business that cannot be identified is hard to recommend.

Then show first-hand knowledge. Describe how you actually do the work, what goes wrong, and what you advise against. A physiotherapist who writes "we see most desk workers three weeks after the pain starts, which is two weeks too late" is demonstrating experience in a way no generic article can. Add the name and role of the person who wrote or reviewed the page, the date it was last reviewed, and links to the sources behind any claim that is not your own observation.

Step 4: Which structured data helps AI search?

Structured data is a block of code, usually JSON-LD, that states facts about a page in a fixed vocabulary from Schema.org. It does not guarantee a citation and it is not a ranking trick. What it does is remove doubt: this is an organisation, this is its address, this page is an article published on this date, these are the questions it answers.

For a service business the useful types are few. Organization or LocalBusiness for who you are. Service for what you offer and where. Article or BlogPosting for guides, with the author, publisher and dates. BreadcrumbList for where the page sits. FAQPage for a list of questions and answers that is visible on the page.

Two cautions. First, the markup must describe what a visitor can see. Marking up questions that are not on the page, or reviews you wrote about yourself, breaks Google's guidelines. Second, do not expect a visual reward for every type. In 2023 Google limited FAQ rich results to well-known government and health websites, so for most businesses FAQPage markup is about clarity for machines, not about extra space in the results.

Step 5: Why do mentions on other websites matter?

An assistant answering "who can help with this in my city" is assembling a recommendation, and it is more comfortable repeating a name that several independent sources agree on. Your own website is one source. The others are what a careful person would check: your Google Business Profile, reviews, directory listings, industry associations, local news and articles where you are quoted.

For a local business, a complete and accurate Google Business Profile is the first of these. Keep categories, services, hours and photos current, and reply to reviews. Ask real customers for reviews after a job is finished. Do not buy them or write them yourself: fake engagement is against Google's policies and it tends to read as fake.

Beyond that, earn mentions the slow way. Contribute a useful comment to a trade publication, publish a small piece of original data, speak at a local event, or write a detailed case study a client is happy to share. This is the work of digital PR. One mention in a publication your customers actually read is worth more than fifty directory entries nobody visits.

A real example: what we changed on this website

We apply the same steps to quaventotechnologies.com, so here is what that looks like in practice. You can check each item yourself.

Our robots.txt uses one open rule for all crawlers and blocks only utility paths, so Googlebot, Bingbot, OAI-SearchBot and PerplexityBot can all read the site. Every guide on this blog opens with a short "Quick answer" box, and section headings are written as questions. Each guide ends with a "Sources & further reading" list, and the same sources are included in the article's structured data as citations.

FAQ sections on our service and city pages carry FAQPage markup, and an automated check runs after every build to confirm that the questions in the markup match the questions visible on the page. Our city pages show the date they were last reviewed. We also publish an llms.txt file that lists our main pages in plain text. We treat that file as a convenience for assistants that choose to read it, not as a ranking factor, for the reason given earlier: Google does not use it.

None of this is exotic. It is a set of small, checkable habits. The value comes from applying them to every page instead of a few.

A worked example: a physiotherapy clinic in Pune

To make the steps concrete, take an illustrative case: a two-therapist physiotherapy clinic in Kothrud that wants to be suggested when someone asks an assistant about treatment for back pain from desk work. The clinic is invented for this example, and the plan is the one we would follow.

Week one is access and identity. Confirm that the site is indexed and that nothing blocks search crawlers. Put the clinic's name, address, phone number and hours on every page, and bring the Google Business Profile in line with them. Add a page for each therapist with qualifications and registration details.

Weeks two and three are content. Replace the single "Services" page with one page per condition treated. Each page opens with a direct answer to the question a patient would ask, such as "How many physiotherapy sessions does desk-related back pain usually need?", followed by what the first visit involves, what the therapists see most often, and when a patient should see a doctor instead. Each page names the therapist who reviewed it and the date.

Week four is structure and reputation. Add LocalBusiness, Service, BreadcrumbList and FAQPage markup that mirrors the visible text. Start asking patients for reviews at discharge. Offer a short article on desk ergonomics to a local newspaper or a housing-society newsletter. After that, the work is a monthly check: are the pages still accurate, and is the clinic appearing when the questions are asked?

How do you measure AI search visibility?

Measurement is less tidy than it is for normal search, so use three simple methods together. First, look at referrals in Google Analytics 4. Visits from assistants arrive with sources such as chatgpt.com, perplexity.ai and copilot.microsoft.com. Build a report or channel group for them and watch the trend, along with what those visitors do next. If analytics is not set up cleanly, fix that first with a proper GA4 setup.

Second, use Search Console for Google. Google states that traffic from its AI features is included in the overall Search Console performance data, under the Web search type. It is not broken out separately, so you cannot see AI Overview clicks on their own, but your total visibility for question-style queries is there.

Third, keep a prompt log. Write down ten to twenty questions your customers really ask, and once a month ask them in Google, ChatGPT and Perplexity. Record whether your business is named, whether it is linked, and which other sources appear. Answers vary from one run to the next, so read the log for patterns over several months and not as a precise score.

What mistakes should you avoid?

Most of the damage we see is self-inflicted and easy to prevent.

• Blocking crawlers by accident: an old robots.txt rule or an aggressive firewall setting that hides the site from the products you want to appear in.
• Publishing pages in bulk with nothing new in them: Google's spam policies name scaled content abuse, meaning many pages produced mainly to rank with little value for readers. How the text was produced does not matter. Whether it helps does.
• Hiding the answer: burying the useful sentence under a long introduction, or putting key facts in images, PDFs or tabs that load late.
• Marking up what is not there: FAQ or review markup that does not match visible, genuine content.
• Chasing shortcuts: paying for an "AI ranking" file or tool while the basics of access, clear answers and trust are unfinished.
• Inconsistent business details: different names, addresses or phone numbers across your site, profiles and listings.

Where should you start this month?

Pick the five pages that matter most to your revenue and work through them in order. Check that they are indexed and readable without scripts. Rewrite the opening of each section so the answer comes first. Add the name of a reviewer, a review date and sources. Confirm the structured data matches the page. Then choose one way to earn a mention somewhere your customers already read.

If you want a second pair of eyes, an SEO audit covers crawl access, indexing, content and structured data in one pass, and our AI search optimisation work picks up from there. For businesses that depend on nearby customers, the local SEO checklist is the natural companion to this guide, and SEO cost in India explains how this kind of work is priced.