Help AI search understand your business by publishing useful, accurate service information on pages that can be crawled and read. Explain who you are, where you work, which problems you handle and how a customer can contact you. Those foundations matter more than inventing a special AI brand, hiding important facts in images or buying a promise of recommendations.
An AI answer can draw on search results and other retrieval mechanisms. Your ability to influence it is limited: you can make the information accessible and trustworthy, but you cannot assume that a particular engine will index, cite or recommend it.
Write the answer a customer needs
A homeowner asking who can assess a heat pump in their area needs service fit, coverage and a usable next step. Put those facts in ordinary visible text. Use clear headings and explain exclusions or assessment requirements where they affect the decision.
Keep the company name consistent across visible pages, titles and structured data. For GlassRiver, the public brand is GlassRiver and the offering is local growth systems for home-service businesses. The existing web domain is an address, not a reason to invent another company or product name.
Link your service, location, team and contact information so a reader can check the context. Real proof should be identifiable as proof; conceptual examples should be identifiable as examples. Avoid writing a collection of near-identical pages solely to match many possible questions.
Understand Google's stated requirements
Google's current generative-AI optimization guide emphasizes normal search foundations and useful, accessible information. It says llms.txt is ignored for Google Search, so that file should not be sold as a Google ranking improvement.
Google's AI-feature guidance explains that supporting pages need to be indexed and eligible for a search snippet. Meeting the requirements does not guarantee inclusion. Have the verified Search Console property owner check the setting for inclusion in generative AI features, alongside indexing and snippet eligibility. Use actual property evidence rather than a local build or successful page request.
There is no reason to force every paragraph into a fixed length or repeat exact question wording unnaturally. Make the content complete enough for the decision, with clear sources for factual claims and useful detail from your own business.
Separate search crawlers from training crawlers
OpenAI documents OAI-SearchBot separately from GPTBot. The former supports ChatGPT search; the latter can collect content for model training. Training permission and search visibility are different controls.
Check robots.txt and hosting rules together. A permissive robots file does not help if a firewall blocks the intended crawler. User-triggered retrieval can operate differently from automatic crawling, so inspect the provider's current documentation instead of assuming one bot rule controls every visit.
Keep private or internal material behind actual access controls. Robots exclusions are crawl preferences, not a login system. Public discovery information should not include customer details, credentials or internal operational records.
Offer optional facts without creating competing truth
A short facts page or machine-readable summary can help some readers and tools retrieve the essentials. Maintain it from the same approved information as the website. Do not let an extended AI text file carry old pricing, unsupported capabilities or stale demonstration promises.
Bing documents sitemaps as a URL-discovery aid. Keep sitemap URLs canonical and current, then inspect the verified webmaster account. Submission or retrieval is not proof of a ranking or citation.
Measure what you can observe
Look for identifiable referral visits, customer-reported discovery and enquiries that reach the team. Not every AI interaction produces a useful referrer, and a citation does not prove a booked job. Keep unknown sources visible instead of assigning them to the channel you hope is growing.
Use the local SEO checklist for the broader foundation and the enquiry-to-job map for conversion evidence. See GlassRiver's approach. Discuss discovery readiness with the actual pages, crawler access and operator evidence you want to improve.