Visibility in ChatGPT and Google AI

Start with questions customers ask

Collect questions from conversations, enquiries and your everyday work. Choose questions you can answer concretely: the work you perform, the area you serve and what someone needs to provide before a visit. Put the short answer first, then add the necessary detail. Avoid pages that merely repeat a popular question without resolving anything. A visitor should be able to use the information without contacting you just to understand basic arrangements. Include relevant limitations rather than leaving every practical condition for a later conversation.

Make your own information verifiable

Use consistent business details and identify who provides the service. Add original explanations of your process, using examples you have permission to share. Include dates and conditions when they change the meaning. Verify prices, qualifications and comparisons before publication. Do not let an AI tool invent sources, reviews or experience. If a fact has not been confirmed, leave it out or explain the uncertainty. A long page of unverifiable assertions is not a better foundation than a short, accurate answer.

• Match the business name, service area and contact details across pages.
• Support specific claims with real information.
• Remove obsolete conditions and invented examples.

Check where the information is published

The Flova pilot under /sites carries noindex. For search visibility, work from a separately verified production environment. Google ties eligibility for its AI search features to indexing and snippet eligibility. Have your maintainer check the final URL, access and relevant settings. A public pilot link or downloaded ZIP does not establish these conditions. Also decide who updates the production copy later. A correction in your working document only helps visitors once it reaches the right public page. Include that responsibility in your handover.

• Use the real production URL for checks.
• Assign an owner for technical configuration.
• Compare published content with the verified source.

Separate search from model training

OpenAI identifies OAI-SearchBot for search and GPTBot for possible model training; their controls are independent. Ask your maintainer to record an intentional choice rather than treating every AI crawler as one category. Give people and bots the same honest content. Avoid cloaking, hidden recommendations and invented AI schemas. Do not treat llms.txt as a magic priority rule. Verify crawler changes on the actual host and document their purpose so that the next maintainer understands the decision instead of undoing it blindly.

Observe outcomes without promises

Record the practical questions your site answers and the information customers still need. Review where visits come from and document any AI mentions with a date and context. One answer is not a stable position: wording and sources can vary. Avoid creating dozens of near-identical pages to force recommendations. Google prohibits deceptive manipulation and scaled abuse. Improve a demonstrably unclear answer instead. When reporting results internally, describe what you actually observed without turning that observation into a promise of future citations or guaranteed recommendation placement.

• Record the question, date and observed reference.
• Check that the referenced page remains accurate.
• Choose improvements based on real information needs.