Start with a question a customer would really ask
AI-search visibility is an observation to document, not a position a business can buy or guarantee. Choose a few questions connected to the work you actually want: a customer comparing services, checking an offer or finding a provider in a place you serve.
For example, a hypothetical Virginia contractor might inspect a question about its core service in its real service area. A broad prompt such as “best company anywhere” says little about the next customer's decision. Record the exact question before testing so a later comparison has a usable starting point.
Check the page and the next step
Open the page that should answer the question. Can a visitor understand the service, its conditions and the places the business actually serves? Can they find the relevant contact route? Is the evidence for a consequential claim visible or linked?
Fix missing or contradictory facts before changing the writing format. A customer needs a clear answer, useful qualifications and a next step. A different heading style or another FAQ block does not make an inaccurate page useful.
Keep search access separate from training choices
OpenAI's crawler documentation distinguishes OAI-SearchBot, which supports ChatGPT search, from GPTBot, which concerns model training. Those settings are independent. Allowing training is not a prerequisite for search visibility. ChatGPT-User handles some user-initiated visits and is not the control for automatic search crawling.
Ask the site operator to inspect the applicable crawler controls and hosting restrictions. A crawl log can show a request; it does not prove that the page was used in an answer. Keep the record of access checks separate from observations of an answer or a referral visit.
Use Google's actual eligibility rules
Google says existing SEO practices apply to its generative AI features. The page must be indexed and snippet-eligible, and the site must be included in Search generative AI features through Search Console. Ask the property operator to verify the setting. Eligibility does not guarantee selection.
Check the page that serves the task instead of creating another page for every wording of the question. Link it from a relevant part of the site and keep important information available as text. The schema guide explains how structured data fits without treating it as a citation switch.
Keep a small observation log
For each check, save the date, platform, exact query, any location or account context you can observe, the answer and the destination of any cited link. Distinguish a brand mention from a link to the business and a link to a third-party directory. They are different observations.
Use the same questions for a later check, but retain the conditions and limitations. Answers can vary. One appearance is evidence of that appearance; one absence does not prove the business can never appear. Avoid converting a small set of manual checks into a claimed market share.
- Answer observation: what appeared for the recorded question at that time?
- Referral observation: did an attributable visit reach the site, where tracking permits?
- Business observation: did a qualified inquiry follow, and what can the available evidence actually connect?
Choose the next edit from the finding
If the answer repeats an obsolete service detail, correct the authoritative business information and the affected website pages. If a useful page is inaccessible, investigate that access problem. If the current page already answers the question clearly, more words may add little. Read the page before deciding it needs a replacement.
With property access, check Google’s Generative AI performance report for Search. It reports impressions for AI Overviews and AI Mode, not branded query volume across AI tools; insufficient impressions can leave the report unavailable. Its data is included in Web performance data, so do not label all Web traffic as AI traffic. Keep manual answer checks, attributable referrals and inquiry records separate.
Send a project inquiry with the business URL, the questions that matter and any existing observations. A useful AI Search scope starts with those findings. It should state the work, the evidence to keep and the reporting limits without promising a citation deadline.