VisibleShift guide · 11 October 2026
Start with the information your customers need, the technical foundations that make it accessible and a way to judge whether your changes are useful.
When a competitor appears in an AI answer and your business does not, it is tempting to look for a special fix. Before changing the website, ask a more useful question: what exactly is missing, and what evidence supports that conclusion?
This guide sets out a practical order of work for a small business. It is a framework for investigation and prioritisation, rather than a formula for guaranteed recommendations.
1. Define the discovery problem
Write down the customer, the service, the location if relevant, and the decision you want to support. “We need better AI visibility” is too broad. “People comparing local bookkeeping providers cannot find clear information about our monthly support” gives you something concrete to investigate.
Separate three questions: whether your business is mentioned, whether the answer describes it accurately, and whether a person can reach a useful page. These are different observations and may call for different actions.
2. Check access before adding content
Google says foundational SEO remains relevant to its generative AI search experiences. Its guidance highlights crawlability, indexing, useful content and a clear technical structure. Meeting the requirements does not guarantee inclusion. Read Google’s guidance.
For your own review, record whether important pages are public, reachable through internal links and intended to be indexed. Ask whoever manages the website to explain any access restrictions before changing them. A deliberate privacy control is different from an accidental technical barrier.
3. Make the business easy to understand
Read your key pages as if you had never heard of the company. Can you identify the services, the audience, the area served and the next useful step? Is the description consistent with your public business profiles?
Create a short list of missing customer information. Examples might include what is included in a service, who it suits, typical constraints and how the process works. Use genuine experience to answer those questions. A clear, specific explanation is more useful to a prospective customer than a page padded with search terminology.
4. Treat platforms separately
OpenAI distinguishes OAI-SearchBot, used for ChatGPT search, from GPTBot, which relates to potential model training. The controls are independent. Allowing search access and allowing training are separate choices. Read OpenAI’s crawler documentation.
Do not assume that a recommendation for one platform applies to every other system. Keep a record of the platform, the question, the date and any visible sources when examining answers. Treat a single response as an observation, then look for a repeatable pattern before committing time or money.
5. Choose one improvement and a review point
Turn the review into a small action plan. For each proposed change, write down the problem, the evidence, the owner and the expected benefit. Start with a change you can explain clearly, such as correcting an inaccurate service description or making an important page easier to find.
Decide how you will review it. You might check whether the information is now accurate, whether visitors reach the relevant page, or whether enquiries are better matched to the service. Record other changes made during the same period so you do not attribute every movement to one intervention.
What can wait?
Google’s guidance says it does not use llms.txt to improve Search visibility and does not require special schema for generative AI search. That is a Google-specific statement, not proof about every AI system. See Google’s mythbusting guidance.
A useful priority test is: does this proposal solve a demonstrated problem, help a customer or make the business information more reliable? If the only justification is that it sounds like an AI optimisation tactic, ask for better evidence before giving it priority.
A manageable starting point
Choose one important service and review its discovery journey from question to useful information. Document what you find, make a focused improvement and return to the same questions later. That creates a practical learning cycle without pretending you can control what every search system will recommend.
Explore the VisibleShift approach or read about VisibleShift.