AI search has brought a new vocabulary: AEO, GEO, AI SEO, llms.txt and “AI-ready websites”. For a business owner, it can sound as though a completely new technical strategy has become necessary overnight.

Google’s own guidance is considerably less dramatic. There are changes worth understanding, but a new label on a proposal does not establish that the work is useful.

This article is primarily about Google Search, including AI Overviews and AI Mode, using guidance reviewed in September 2026. Other AI platforms may crawl, select and present information differently. Advice about Google should not be presented as a universal rule for every service.

What people mean by AI search

Google’s AI Overviews present a generated summary for some searches, with links to supporting information. AI Mode offers a more conversational experience for exploring questions and comparisons. The reader may encounter part of an answer before deciding which website to visit.

AEO usually means answer engine optimisation; GEO means generative engine optimisation. These terms describe work aimed at visibility in generated answers. Their use is not evidence that a supplier has access to a special system or can guarantee a place in an answer.

In its guide to optimising for generative AI features, Google says that familiar SEO practices remain relevant. It does not prescribe a separate AEO or GEO technical layer for Google Search.

The entry requirements are familiar

Google’s AI features documentation states that a page must be indexed and eligible to appear with a snippet to be considered as a supporting link. There are no additional technical requirements for AI Overviews or AI Mode, and meeting the requirements does not guarantee inclusion.

Start by checking whether important public pages can actually be reached. Content behind a login, an accidental crawling restriction or a broken internal link needs attention before a discussion about AI visibility. Check the live page, its indexability and the route visitors take to it.

An important service explanation should be available as readable text. A diagram can help, but it should not be the only place where the business describes what it provides. Visitors also need a clear path from that explanation to relevant evidence and contact information.

Ask what your page adds to the answer

The more useful strategic question is what someone gains by reading your page. A generic description of a service may be accurate while adding very little that distinguishes the business or helps a customer decide.

Consider a service page that says only that work is professional, reliable and tailored. A stronger version could explain what the service includes, which constraints affect the scope, what information is needed to quote and which requirements fall outside the offer. These details give the reader something specific to evaluate.

First-hand knowledge can appear in modest forms. Explain a real process, describe a limitation that buyers often need to understand, or show an actual piece of work with permission. It need not become an elaborate case study, and it should never depend on invented results.

Google’s people-first content guidance provides useful questions about originality, expertise and whether a page leaves its reader better informed. Use those questions as an editorial review, rather than a formula for repeating particular words.

For an established business, a carefully maintained set of useful pages can be a better investment than publishing every topic adjacent to its sector. Start with the questions that affect whether someone can sensibly choose the service. Give those answers an owner who can keep them accurate.

Clear structure helps without becoming a secret technique

Use descriptive headings and sensible internal links so readers can follow the subject. A heading such as “What the survey includes” gives more context than “Our approach” when the section explains a specific deliverable. That is useful editing regardless of where the visitor came from.

Semantic HTML supports clarity and accessibility. Google does not require perfectly semantic HTML for its AI features, and it should not be sold as a hidden mechanism for gaining citations. Use meaningful elements because they describe the page properly.

Follow the same principle with related content. Link when another page answers the next question, not because every paragraph needs an SEO link. A service page, relevant project and clear enquiry route should work together as a coherent journey.

No special schema or llms.txt file is required

Structured data remains useful where it accurately describes the page and supports an applicable search feature. There is no special “AI schema” required for Google’s generative answers. Adding more markup does not make unsupported claims more credible.

Google Search does not require or use llms.txt. Its current guidance says that maintaining such a file neither helps nor harms Google Search visibility or rankings. A file may be relevant to another service that uses it, but that is a separate, service-specific reason to maintain one.

When a proposal includes an AI-readiness file or markup package, ask what documented requirement it addresses, which platform uses it and how the work will be verified. An answer should identify a concrete purpose, rather than rely on the possibility that it might matter someday.

The visit still needs to work

Being discovered is only one part of the job. A visitor still needs to read the page, understand the offer and use the site on their device. Slow loading, unstable layouts and an awkward form remain problems even if the first introduction happened inside an AI answer.

Review Core Web Vitals and real website performance alongside mobile usability. Check accessibility as part of the same experience, including keyboard access, readable text and properly labelled controls.

Use images and video where they explain something the text cannot show as effectively: a genuine project, a product detail or a process that is easier to demonstrate. Do not add media just to make a page appear more comprehensive. It should have a useful role and be delivered carefully.

Accurate company information, specific service descriptions and evidence that matches the claims contribute to a trustworthy website. Those details deserve attention before another badge or a claim that the site is ready for the future.

Publish useful work, then assess what it achieves

Generating hundreds of low-value articles creates more pages to check, maintain and keep consistent. It does not automatically create expertise or make a business more relevant. Generic AI-generated copy has no inherent advantage over equally generic copy written another way.

A sensible review asks whether the content is accurate, original enough to be useful and appropriate to the business publishing it. Check facts, remove unsupported claims and improve pages when the service changes. Do not turn every variation of a question into another thin article.

Assess enquiries and useful visits alongside visibility. Being mentioned in an answer is not the same as a customer understanding the service, and nobody can responsibly guarantee a citation. Ask suppliers to distinguish measured results from assumptions about how answers are selected.

AI search makes good website fundamentals more important, not obsolete. The practical response is a clear, useful site that deserves the visit when someone arrives, with changes justified by evidence rather than new terminology alone.