How to Get Your Business Cited by AI Search Engines

AI answer engines do not rank pages; they decide whether to trust and name you. That changes what "optimising" means. We have done this work on our own site, so this is written from the inside.

Where AI answers come from

Google's AI Overviews and AI Mode draw on Google's index. ChatGPT search draws on Bing's index. Perplexity and Anthropic's Claude crawl the web directly with their own bots. Each then reads candidate pages and decides which sources to cite. So three things have to be true: they must be allowed in, they must be able to read the page, and they must have a reason to believe you.

Step one: let them in

Check your robots.txt. Many sites block AI crawlers by default, often via a CMS plugin or a hosting template, without anyone deciding to. The crawlers that matter for being cited (as opposed to being used for training) include OAI-SearchBot (ChatGPT search), PerplexityBot, ClaudeBot and Claude-SearchBot, and Google-Extended for Gemini. If you want to be cited, allow them explicitly. It is a business decision - some publishers choose to block - but for a business that wants customers to find it, blocking is self-harm.

Step two: make the page readable without JavaScript

Most AI crawlers do not execute JavaScript. If your content only appears after scripts run - common with certain site builders and single-page frameworks - the crawler sees an empty shell. Server-rendered HTML with the content in it is the fix. View the page source in a browser: if the text is there, you are fine.

Step three: answer-shaped content

Engines quote passages that answer a question directly. Content that works: a visible FAQ with complete answers, definitions, specific numbers with units and timeframes, comparisons in tables, and headings that are the questions people ask. Content that does not: vague brand language, claims with no figures, walls of text with no structure. The local SEO guide on this site is written that way deliberately.

Step four: llms.txt

An emerging convention: a plain-text file at /llms.txt that describes what the site is, who is behind it, and where the substantive pages are, in Markdown a model can read cheaply. Some engines fetch it; others will. It costs an hour and it is a signal that the site is maintained by people who understand how it is read. Ours is at /llms.txt, with a full-text companion at /llms-full.txt.

Step five: entity clarity

An engine has to be sure which "Acme Ltd" it is looking at. Use one name, one address, one phone, one description, everywhere: your site, schema markup, Google Business Profile, LinkedIn, directories, Companies House. Link your profiles from the site (schema sameAs) and link the site from each profile. Put the legal entity name and company number somewhere visible. Named people with real bios and profile pages help enormously - an organisation with no humans in it is harder to trust.

Step six: the hard part - corroboration

This is where most small businesses stall. Answer engines weight third-party evidence heavily. If the only site that says you exist is your own, you will rarely be cited, however good the content. What counts:

• Reviews on Google, Trustpilot, Clutch or the platform for your industry.
• Directory profiles with consistent details: the industry directories appear in AI answers for "best X in Manchester" queries constantly.
• Press and mentions - local business press, trade publications, event listings.
• Links from clients, suppliers, partners and networks.
• Case studies with real figures and, ideally, named clients who can be checked.

None of this is new; it is what search engines have valued for years. AI engines simply weight it more and are less forgiving of its absence.

Step seven: freshness and measurement

Date your content, update it, and show the date. Add an RSS feed so aggregators and crawlers notice new articles. Use IndexNow to push changes to Bing (and therefore ChatGPT) instantly. Then measure: Search Console shows AI Overview impressions; Bing Webmaster Tools shows its crawl; server logs show which AI bots visit and what they read.

A realistic timeline

How this differs from ordinary SEO

Traditional search ranks ten links and lets the reader choose. An AI answer synthesises and cites a few sources, so there is no "position seven". You are either used or you are not. That makes two things more important than they were: being the clearest, most complete answer to a specific question (rather than one of many adequate answers), and being trusted enough that the engine is willing to put your name in its response. Ranking factors have not gone away - the engines draw on the same indexes - but the threshold for being cited is higher than the threshold for being listed.

Content that gets quoted

Engines lift passages. A passage gets lifted when it stands alone: a question as a heading, a direct answer in the first sentence, the supporting detail after. Numbers with context ("8 to 16 weeks from signed brief to launch") are quoted far more than adjectives. Comparisons in tables are quoted. Definitions are quoted. Long preambles, marketing language and claims without figures are skipped. If you write the page so that any two paragraphs could be pulled out and still make sense, you have written it for AI engines.

Checking your own site in ten minutes

1. Open /robots.txt and look for the AI user agents. Are they allowed?
2. View the source of your main service page. Is the copy in the HTML, or does it arrive by script?
3. Search your business name in ChatGPT, Perplexity and Google with AI Overviews. What do they say, and what do they cite?
4. Search "your service + your town". Are you named? If not, who is, and what do they have that you do not - usually reviews and directory profiles.
5. Check your details on the first three directories you find. Do they match the site exactly?

Those five checks find the problem for most businesses. The fix for one to two is technical and quick. The fix for three to five is the corroboration work in the local SEO guide, and it takes months - which is why it is worth starting before the technical work is even finished.

Steps one to five are a week of work for a well-built site. Step six is months, and it is the one that matters. A business that does the technical half and ignores corroboration will have a perfect site that no engine names. We build the technical half into every site we deliver and run the rest as part of our SEO and digital marketing work - ask us where your site stands.