SEO and AI search: what should you actually do first?

SHORT ANSWER

Start with a real customer question, a useful page and a way to evaluate your work. That is a firmer foundation than a promise to crack an algorithm.

Svenska Digitalakademin 6 November 2026

White metal cladding with sharp diagonal lines against a pale sky. — Photo · Scott Webb

Choose a question that affects a decision

Imagine a business that installs a product. A potential customer wants to know whether it fits their situation, what the work involves and what affects the cost. Start with that question. Check whether one page gives an understandable answer or leaves the visitor searching through slogans and contact buttons.

A question becomes useful when you can connect it to a particular audience and a next step. Write down the decision the page should support. It then becomes easier to choose the explanation, reference material and links a visitor is likely to need.

Make the page discoverable and understandable

Review the heading, body copy and internal links. Important information should be readable as text. Images can demonstrate a result or explain a process, but should not carry the entire answer. A technical review also needs to establish whether the page can be fetched and whether it is intended for indexing.

Google states that ordinary SEO principles apply to its AI features in Search. Special AI schema is not required. Work with readable content, accessibility and information consistent with the visible page.

Define a change that can be assessed

A reasonable first proposal might be to explain when the service is suitable, show an original case and link from a relevant overview page. State why the change is needed and what evidence supports it. Make missing data explicit. A hypothesis is a basis for work rather than an established result.

Choose how to evaluate the change. Relevant queries, visits to the page and meaningful visitor actions can each show part of the picture. Document completed changes so you can later discuss what may have influenced the outcome.

Set a clear boundary for the review

Decide which pages and questions belong in the review before beginning. An assessment of one service page should not turn into a promise about the growth of the whole business. A well-scoped review can be handed to a colleague who understands both its coverage and the work still needed.

Finish with the first action, the role responsible and the checks needed before publication. State when a change requires developer support. A recommendation becomes useful when someone can assess its value and what it will take to implement.

Treat an AI answer as an observation

When reviewing AI search, record the query, service, date and links actually shown. One answer does not establish how often a business appears or prove why it was mentioned. Answers can vary and services do not expose every part of the process.

A useful first work sample is a focused visibility review with your observations, sources and a prioritised improvement. It shows that you can make a reasoned choice. No course can reliably promise that a particular change will produce a specific ranking or AI mention.