Website check
SEO & GEO
Its own profile and its own report, deliberately separate from security and privacy. These are different trades and different people: someone who wants to know why their site is not being found should not have to scroll past cookie findings. Severity here does not mean danger but impact on visibility — "critical" means the page does not appear at all.
What gets checked
24 individual checks in a dedicated discoverability run
-
Indexability
noindex via meta tag or X-Robots-Tag, and a robots.txt disallowing everything — the two mistakes that make a site entirely invisible and are routinely missed at launch.
-
Local discoverability
For location-based businesses the single most important lever: LocalBusiness markup with address, coordinates, opening hours and service area, postal address and phone number in visible text, and the location named in the title and H1. Naming the location nowhere means competing with the whole country for no reason.
-
Title, description and headings
Present, sensibly sized, no meaningless titles such as "Home". H1 present and unambiguous, no skipped levels in the outline.
-
Canonical address and sitemap
Canonical missing, diverging, or pointing at a foreign domain. XML sitemap present, populated, and referenced from robots.txt.
-
Sharing and structured data
Open Graph and schema.org markup. For local businesses the lever with the best ratio of effort to effect.
-
Images and technical basics
Missing alternative texts, images loaded far larger than needed, lazy loading. Plus language declaration, mobile viewport and favicon.
-
Content and loading
Amount of text, time to first byte, load time and bytes transferred.
-
Visibility in AI answer engines (GEO)
Whether robots.txt blocks GPTBot, ClaudeBot, PerplexityBot and others — a legitimate decision, but one that should be taken deliberately rather than inherited from a plugin default. Plus question-and-answer markup and whether the page carries quotable statements at all.
-
llms.txt and llms-full.txt
Both are fetched and assessed together. If they are absent, that appears as a note in the report — explicitly not as a defect, since no evaluation by the major providers has been demonstrated to date. The full version is only recommended for sites with substantial text. A near-empty llms.txt, on the other hand, is flagged: a file that says nothing is worse than no file.
Standards and references
- Google Search Essentials and the structured data documentation
- schema.org, in particular LocalBusiness for location-based businesses
- The published crawler identifiers of OpenAI, Anthropic, Perplexity and Google
- llms.txt — a convention, not a standard, and treated as one in the report
What the report contains
A separate report, apart from security and privacy, sorted by impact on visibility. The GEO section is explicitly marked as carrying low confidence.
Where the audit reaches its limit
The homepage is what gets audited. Beyond that, nobody can currently measure reliably why an answer engine cites a source. Only the necessary condition is testable: that nothing technical stands in the way. Not covered is everything that can only be judged across several pages and over time — content quality, inbound links, competition for search terms, actual placements. No guarantee of rankings can be derived from technical improvements, and the report says exactly that.
Frequently asked questions
What prospective clients most often want to know before commissioning an audit.
What is GEO and how does it differ from SEO?
GEO stands for generative engine optimisation: making sure AI answer systems such as ChatGPT, Claude, Perplexity or Google AI overviews can fetch your content, parse it correctly and name you as a source. SEO targets a position in a result list; GEO targets inclusion in an answer. Technically the two overlap a great deal — clean HTML, content without JavaScript, clear structure — but GEO has its own requirements: access rights for AI crawlers, self-contained quotable paragraphs, evidenced and dated statements.
Should we allow AI crawlers or block them?
That is a business decision, not a technical one. If you want to be found through AI answers, you have to grant access — blocked content does not get cited. If you do not want your content in training data, you block and forgo that visibility. Both are defensible. What is not defensible is the third variant, the one we find most often: blocking without knowing it, because a copied robots.txt or a bot protection made the decision silently.
What is llms.txt and do we need it?
llms.txt is a file in the site root telling a language model, briefly, what the site is about and which addresses carry the important content. It is a convention, not a standard, and no search engine requires it. The effort is small and the downside zero, so we check it — but the fundamentals matter more: that content is in the HTML without JavaScript, that headings form a real hierarchy, and that statements are quotable on their own.
Can you guarantee a ranking or a mention in AI answers?
No, and anyone guaranteeing that is selling something they cannot deliver. Rankings emerge in systems nobody can reach from outside, and generative answers vary from query to query. What can be guaranteed is the opposite: that nothing technical stands in the way — that indexing is allowed, content is parsable, crawlers are not blocked and the page loads fast enough to be considered at all.
Does this replace an SEO agency?
No. We audit the technical foundation and say what is broken — we do not write content, research keywords or build links. For an agency or an in-house team the report is where they start: an evidenced list of what is currently holding the content back, instead of a guess.