Assessment
A short first look
You want to know whether the question is worth pursuing.
Three documented findings and three actions to take. Short, and with no commitment attached.
AI visibility (AEO / GEO / SEO)
Be found,and described correctly.
AI visibility is the work of making a business findable, readable and accurately described in AI assistants and search. It is sometimes called GEO, for generative engine optimisation. We observe first, document what we find and recommend only what the evidence supports.
Assessment → diagnosis → 30-day pilot → monthly follow-up
01 / Four formats
From a first look to ongoing follow-up.
Each format has defined content and exclusions, agreed before work starts.
Diagnosis
A verified inventory
You need to know what assistants and search see on your site today.
An inventory of up to ten pages and two agreed competitors, the questions your prospects ask, and a prioritised backlog: evidence, expected impact, effort, owner, dependency and acceptance test.
30-day pilot
A delivery window
You want to try improvements and check their effect.
Days 1–5: diagnosis and baseline. Days 6–15: drafts. Days 16–23: integration and acceptance. Days 24–30: the same checks replayed, and a before/after review. Feedback is gathered in two rounds.
Monthly follow-up
A routine kept in writing
Visibility drifts as assistants, competitors and your own site change.
A versioned panel of questions, a monthly review and a log of every intervention.
02 / Method
Observe, document, then explain.
The order matters. We look at what assistants actually answer before proposing causes.
- 01
Frame the scope
One offer, one market, one language. Three buying intentions: choosing a provider, understanding the method, getting it done. Questions with the brand name are kept apart from questions without it.
What you receive: Scope and question panel - 02
Photograph the technical state
Fetch the pages and record what is actually served: initial HTML, title and main heading, meta tags, canonical, parseable structured data, robots.txt, sitemap and indexing headers, with a fingerprint of each page.
What you receive: Dated technical snapshot - 03
Observe assistant answers
Ask the same questions, in the same language, with the same tool, mode and country, in a session with no history. Record what is said before explaining why.
What you receive: Dated log of answers - 04
Build the backlog
Each line states the evidence, the hypothesis, the effort, the owner and the acceptance criterion.
What you receive: Prioritised backlog - 05
Keep an intervention log
Record the date, the URL, the problem observed, the hypothesis, the change, the check and the next review.
What you receive: Written intervention log - 06
Replay identically
At the end of the pilot, repeat the same checks. Any break in comparability, such as a changed model or a missing answer, is documented.
What you receive: Before/after review
03 / Rules of proof
What we will and will not claim.
These rules apply to every report, so that findings can be checked and nothing is overstated.
An observation, not a score
We do not give a global score. A small sample is described as a sample, and we report the number of answers, failures and invalid responses.
Mention is not citation
We separate a mention of the brand from a citation of a URL on your domain.
Absence does not prove a cause
An answer without your brand is a finding, not an explanation. A before/after difference does not show that our changes caused it.
Four levels, kept apart
Visibility, visits, qualified enquiries and sales are different things. Sales are outside our scope.
Not measured stays blank
Where we have no access, indexing, traffic, conversions and AI citations are marked not measured. We never invent a baseline.
Technical signals have limits
robots.txt is a declaration, not proof of a real visit. A request with a chosen user agent can be simulated, and an isolated non-200 status does not prove invisibility.
Sources are checked
Claims about crawlers and search features are checked against primary sources, such as search engines and AI providers, before we advise or configure anything. Search and training crawlers are treated separately.
Nothing bought, nothing invented
No purchased reviews, mentions or links, and no invented testimonials or metrics.
Practical questions
Can you guarantee that assistants will recommend us?
No. Presence observed on a date is not an earned position, and no assistant recommendation can be guaranteed. Thirty days is a delivery window, not a promise of search results.
What is not included?
Implementation beyond the pilot, access to private data without agreement, general SEO unrelated to AI visibility, purchased links, paid tracking tools and guaranteed results.
Is this the same as SEO?
It builds on SEO foundations: clean structure, correct metadata, crawlable pages. AI visibility adds the observation of assistant answers and the clarity of the facts about your business. It does not replace SEO.
Do you rely on llms.txt or special markup?
llms.txt is not required for Google, and structured data must match the visible content without any promise of extraction. We recommend them only where they serve a documented purpose.
Other services
Start with a first look
How do assistants describe your business today?
Tell us about your business, your market and the questions your prospects ask. That is enough to scope an assessment.
Discuss AI visibilityPresence observed on a date is not an earned position.