SEO has not been replaced by a second discipline called “AI SEO.” Search results, AI summaries and conversational research may present information differently, but they still depend on accessible pages and useful sources. The change for an agency is in how it plans evidence, measures discovery and explains the limits of what it can promise.
The Google Search guidance for generative AI features explicitly says foundational SEO remains relevant. A page generally needs to be indexed and eligible for a snippet to appear in Google’s AI search features. There is no special schema that buys inclusion. That is a more useful starting point than the claim that “old SEO is dead.”
Search visibility, today.
Keep the foundation your client already needs
Start with the client’s important services and the pages that represent them. Check whether those pages can be crawled, rendered, indexed and reached through internal links. Review titles, canonical choices, redirects, mobile usability and the actual page copy. A website with inaccessible service pages will not be rescued by an AI-flavored FAQ.
Keywords still help reveal what people ask and how their vocabulary differs from the agency’s. They should inform the page’s job, not become a quota of repeated phrases. Search Console queries, sales calls, proposals and support questions can show whether the site answers the work the client actually wants.
For example, a studio may sell a restaurant website. A page that only says “beautiful digital experiences” omits the decisions a restaurant owner needs to make: menus, bookings, languages, opening hours, ownership of accounts and how updates happen. Those specifics make the page more useful to a person and easier to interpret in search.
Add evidence that belongs to this client
AI can paraphrase generic advice cheaply. The client’s real process, constraints and observations are harder to replace. Ask what the team has learned from implementations, customer questions and the trade-offs it has made. Then publish the parts it can substantiate: a decision guide, a comparison with clear criteria, a documented process, or a case with permission and honest limits.
This does not require turning every paragraph into a 40-word answer. Google says there is no required “chunking” format or ideal page length for AI search. Write a clear answer near the relevant question, then give the detail a buyer needs to make a decision. Use ordinary headings and descriptive links because readers benefit from them, not because a template claims to unlock AI citations.
If a client wants to cite an outcome, ask for the baseline, time period, measurement method and permission to publish it. A plausible but unverified percentage is weaker than a precise account of what changed and why.
Measure several outcomes without mixing them up
Ranking positions and organic visits are still useful, but they describe only part of discovery. In the client report, separate four questions:
- Can priority pages appear? Check indexing, queries and impressions in Google Search Console and Bing Webmaster Tools.
- Where are people encountering the brand? Review conventional search data and, where available, AI citation or visibility reports. Bing’s AI Performance report identifies cited pages and related queries in supported Microsoft experiences; it is not a complete view of every assistant.
- What happens after a visit? In analytics, distinguish qualified enquiries from raw traffic and check that forms, calls and attribution actually work.
- What has the team learned? Keep a short record of pages improved, questions answered, evidence added and assumptions disproved.
An AI mention is not a sale, and the absence of an observable mention in one tool does not prove a page was never used elsewhere. Similarly, a traditional ranking chart does not tell the whole story of a multi-step buying decision. Report the data each platform actually exposes.
What I would do with an agency client first
In the first working session, I would select three to five commercially important pages and trace them from crawlability to enquiry. Then I would identify the questions that keep coming up in briefs or sales conversations, and decide whether existing pages can answer them or whether a new page has a distinct role. Only after that would I propose a content calendar.
The result is an SEO plan the agency can explain and deliver: technical access, useful answers, credible evidence, deliberate internal links and measurement. It works as a service an agency can bring to its client without promising a citation in Google AI Mode, Copilot, ChatGPT or another product.
If your agency needs someone to inspect the site, make the technical changes and help you present the priorities to your client, that is the role I take as a technical SEO partner for agencies. We can define the scope and the client-facing arrangement together.