Skip to content

SEO for GEO

AI answers are built on what search engines can already read.

Every engine runs a live web search before it answers. If a page cannot be crawled or understood, nothing else you do will get it quoted. This layer decides whether you are eligible at all.

One page load and a PageSpeed Insights lookup. No sign-up, nothing kept beyond the domain, and no cost to run, which is why there is no gate on it.

What the tester checks

Four groups, in the order an engine hits them. Every item is binary or a measured number against a published threshold, so nothing here needs a keyword from you and nothing is a score out of a hundred.

Reach

Eligibility

Whether an engine can use the page at all. HTTP status and the final URL after redirects, noindex in the meta tag or the header, whether the canonical is self-referencing, cross-domain or absent, and whether robots.txt disallows the path you just tested.

Agents

AI crawler access

Which named answer engines your robots.txt bars. GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot, Google-Extended, CCBot and Bytespider are each checked by name. A large share of sites block some of them without knowing, usually inherited from a security plugin or a CDN default.

Speed

Core Web Vitals

LCP, INP and CLS against Google's own thresholds of 2.5 seconds, 200 milliseconds and 0.1. Field data from real visitors where the origin has it, lab data as the fallback, and the report states which one it is showing. Nothing else gets a pass or a fail.

Quote

Answer-readiness

Whether there is anything on the page worth quoting. Schema parsed properly including nested graphs, FAQPage presence, question-shaped headings, and the gap where a page has several question headings and no FAQPage. Then title and meta description length, a single H1, heading order, sitemap presence and whether robots.txt references it, missing image alt text, thin copy, comparison tables, pricing, and stats an engine could cite.

What a page test cannot tell you

Two of the most expensive problems are invisible to any single-page tool, because they are properties of a site rather than of a page.

Site

Information architecture

Internal linking decides which pages an engine treats as authoritative. Most finance sites scatter authority across near-duplicate service pages competing with each other.

Split

Cannibalisation

Two pages targeting one intent split the signal and neither wins. In AI answers the effect is worse, because the engine has to pick and often picks neither.

It shows you where, not just what

Findings that have a location on the page are marked on a screenshot of it, so a fix can be handed to whoever owns that template without a written description of where to look. Findings with no location, like a missing canonical, are listed above it instead.

Four ways to be absent

Each of these fails for a different reason and needs a different repair. Knowing which one you hit is most of the work.

01Never fetched

Crawled

Blocked, unreachable, or rendered only in JavaScript the crawler did not run.

Fix. Make the page fetchable and server-rendered.

02Fetched, but unclear

Understood

No schema, vague headings, claims buried in prose. The engine cannot tell what you do or who for.

Fix. State it plainly and mark it up.

03Understood, not chosen

Retrieved

You are legible but not relevant to the question asked, or a stronger source answers it better.

Fix. Answer the question the buyer actually asks.

04Read, not named

Cited

The engine used your page but credited someone else, usually one with quotable data or clearer proof.

Fix. Give it something worth quoting.

Questions

Is SEO still worth doing if buyers use AI?
Yes, and more than before. The engines run live web searches to answer questions. If a page cannot be crawled, parsed or understood, it cannot be cited, so the technical layer decides whether you are eligible for an answer at all.
Do I need SEO and GEO as separate engagements?
No. The technical work is part of the fixing work, not an add-on. It is separated here because buyers ask about it separately, not because it is billed separately.
We already have an SEO agency. Does this conflict?
Usually not. The audit identifies what is blocking AI visibility specifically, and a lot of that is work an existing agency can carry out. Some clients take the findings straight to their current team.

The audit checks this layer as well as the answers.

It reports which of these signals are present on your pages and which are missing, alongside the visibility results.