GEO vs SEO: what changes in your agency’s client work
By Johnny Lagneau · Updated September 8, 2026Originally published April 22, 2026
SEO and GEO share much of the same foundation: accessible pages, accurate information, and useful evidence. The difference for an agency is also in what it observes. Search reports describe search performance; an AI visibility audit records the answers produced by a defined set of models and questions.
This guide is about fitting the second kind of work next to the first without creating a confused promise. It is written for an agency that already sells SEO to local businesses.
What stays useful
Everything that makes a page readable by a search engine also makes it readable by a model that searches: crawlable structure, one clear topic per page, accurate facts, real sources, internal links. None of that work is wasted. Google states that its AI features rely on the same signals as search and require no special markup;1 whether the other providers behave the same is not documented by them, so do not assume it.
So the first answer to “do we need a separate GEO programme” is often: not yet. Fix the pages first. The separate part is the measurement, not the foundation.
What you measure differently
Here is where the two diverge. A search report and an AI answer record answer different questions, and the table below keeps them apart on purpose.
| Question | Search report | AI answer record |
|---|---|---|
| What is observed | Impressions, clicks and position for a query, aggregated | The answer a named model gave to a fixed question on a date, in a given configuration |
| Where it comes from | The engine’s own reporting | Your own run, kept as text |
| What it can show | Trend over time, per page and query | Presence, description and recommendation, per model and mode |
| Its limit | Says nothing about AI answers | A sample on a date; models change; no volume attached |
No column above is “better”. They are different instruments. Do not build a score that combines them, and do not publish stars.
What changes in the deliverable
A search deliverable centres on positions and pages. An AI answer deliverable centres on the answer itself: the raw text, whether the business was present, how it was described, whether it was recommended, the source the model cited if any, and the context — model, date, configuration, exact question.
One substitution to avoid: a position in a results page is not interchangeable with a preference in a generated answer. “Ranked third” and “recommended second” are not the same fact, and a client will notice if a report treats them as one.
How to scope the work
Three phases, with a clear owner for each:
- Initial diagnosis. The questions, the run, the record, the findings. Owner: whoever runs the audit.
- Implementation. The pages written, corrected or added. Owner: the agency, on the client’s site.
- Follow-up. The same questions, later, compared with the first run. Owner: the audit provider, on a schedule the client agreed to.
Write the owner into the proposal. The most common confusion in this work is a client who thinks the audit provider will change the website, or an agency that thinks the audit includes a retest it does not.
What not to promise
- A guaranteed ranking or a guaranteed citation. No provider offers either.
- Revenue deduced from a mention. Being named is an observation; a sale is another.
- A timeline for “the model learning about you”. Nobody outside the provider knows when a model’s retained knowledge changes.
- A single visibility score across models and modes. It hides exactly the differences the work exists to find.
How the two reports can work together
Run them in parallel and read them separately. A page that gains impressions in search and starts appearing in AI answers is a useful signal; a page that does one and not the other is a useful question. Neither report explains the other, and neither proves cause. Resist the temptation to compute a return from the pair — there is no honest arithmetic that gets you there.
Two answers to two questions
Use SEO and AI visibility evidence to answer different client questions. Keep the work connected, and keep the measurements distinct. For what the AI side of that looks like in practice, see the guide to GEO for local businesses and how we measure.
1. Google Search Central, “AI features and your website” (developers.google.com/search). Applies to Google’s systems only.
