AI Spy — AI search visibility audits
I spy with my little AI…three competitors, and not you.
Your customers are asking ChatGPT instead of Google. We find out what it tells them about you, and fix it when your name is missing.
CHATGPT · PERPLEXITY · GOOGLE AI · IE
“Which accountancy firm should I use for a small business in Dublin?”
- 01Harbour & GrayNAMED
- 02Rathmines AccountancyNAMED
- 03Cleary Finance PartnersNAMED
- —Your firmABSENT
Illustrative — your audit tests the questions your own customers ask.
01 — The problem
Search used to give ten links. AI gives one answer.
Buyers who once typed a search and worked down a page of results now ask a question and get a short recommendation naming two or three companies. There's no page two. If your name isn't in the answer, you were never in the running — and nothing in your analytics will tell you it happened.
The models build their picture of you from your website, directories, reviews, press and LinkedIn. Where that evidence is thin, inconsistent or out of date, they either skip you or describe you wrong. Both are fixable, but only once you know which is happening.
02 — What we audit
What you get
Fixed scope, fixed fee, and a report written so your own team or your web agency can act on it without me.
- 01
The questions your customers actually ask
We build a list of the real buying questions for your sector, then run every one across ChatGPT, Perplexity and Google's AI answers — recording who gets named, in what order, with dates and screenshots.
- 02
Who's being named instead of you
Your competitors' share of those answers, and what the models say about them that they don't say about you.
- 03
Mentioned versus actually cited
Being named and being trusted are different things. We track whether a model is repeating your reputation second-hand, or pulling from your own site as a source — the second is what you can actually build on.
- 04
What the models can read on your site
Page structure, headings, structured data, and crawlability — whether your services, locations and credentials are stated in a form a machine can extract.
- 05
Whether your facts agree with each other
Your name, address, services, accreditations and claims as they appear on your site, Google Business, directories, LinkedIn and press. Contradictions are one of the most common reasons a model leaves a company out.
- 06
The evidence trail behind you
Reviews, mentions, citations and third-party coverage — the material the models actually draw on when deciding who to recommend.
- 07
Every market you sell into
The same tests, run in each market's own language — including whether AI Overviews are even switched on there, and whether a crawler can find your translated pages at all.
- 08
What to change, in order
A prioritised list separating quick fixes from real technical work, sequenced over 30, 60 and 90 days — then a retest so you can see what moved.
03 — Your score
What we spotted, as one number out of 100.
Every audit ends with an AI Visibility Score out of 100. It isn't a vanity metric — it's built from what the assistants actually said when we asked them, weighted towards the things that decide whether you get recommended.
How the 100 points are made up
- Named in the answer30
How often you appear at all, across your core buying questions.
- Share against competitors20
How you rank when you do appear, and who is taking the space.
- Machine readability15
Whether a model can extract what you do, where, and for whom.
- Consistency of your facts15
Whether your site, directories and profiles agree with each other.
- Evidence behind you10
Reviews, citations and third-party coverage the models draw on.
- Other markets10
Whether the picture holds up in the languages you sell in.
What your score means
- 0–25Invisible
You're not in the conversation. Competitors are being recommended in your place today.
- 26–50At risk — act now
You surface occasionally, usually far down. The gap widens as more buyers switch.
- 51–75Present but beatable
Named for some questions, missing from others. Specific, fixable gaps.
- 76–100Strong — defend it
Consistently recommended. The work now is holding that as competitors catch up.
Scored again after the fixes, so you can see the movement.
04 — Other markets
If you sell abroad, English isn't the test that matters.
The same question asked in German or French routinely returns a different set of companies — and in the EU, Google's AI Overviews may not even be switched on, while ChatGPT, Gemini and Perplexity already are. Almost nobody checks this per market, which is usually why nobody has fixed it.
Sample — medical device components supplier
4 markets
- IE / ENNAMED 2nd
medical device component supplier Ireland
- DE / DENOT NAMED
Zulieferer Medizintechnik Komponenten
- FR / FRNOT NAMED
fournisseur composants dispositifs médicaux
- ES / ESNAMED 4th
proveedor de componentes sanitarios
Illustrative — this is the table your audit fills in with your own markets.
Beyond visibility
Once you know the gap, we can help you close it.
An audit tells you where you stand. If you want practical help getting your own team using AI day to day — for marketing, and for the repetitive work that eats your week — that's a separate, fixed-scope service.
05 — Pricing
Start with the free check. Pay only if it's worth doing.
Everything beyond the check is fixed scope and fixed fee, agreed in writing before any work starts.
Take a peek
Free
Five questions your customers ask, run across the major AI assistants, and a short note on who's being named instead of you. No obligation — I'll tell you if there's nothing here worth paying for.
Full audit
€1,000 from
The complete picture across your markets and languages, with the evidence behind every finding, a prioritised action list over 30, 60 and 90 days, and a retest once the changes are made.
Audit + fix
€2,500 from
The audit, plus the changes carried out alongside your web team — content, structured data and the technical fixes — then retested.
“I have had internal and/or external SEO and, recently, AI search experts/agencies advise me for over 20 years. Search is a specialist area and every expert has their own theory on what and how to drive success. Brendan is the most expert AND practical search expert I have worked with. Not only did he provide our internal approach at Lionbridge Technologies but he advised dozens of strategic F1000 clients on their global search approach repeatedly. If I need search expertise from any one individual, Brendan is the expert I choose.”
KC LINCOLN — CMO, LIONBRIDGE & ENTREPRENEUR
100+
Markets & languages supported
12
Years at Lionbridge (SEO & AI)
3+
Years hands-on GEO/AEO
F1000
Global clients advised
06 — About
Who does the work
Every audit is run and written by Brendan Walsh. Nothing is outsourced, and nothing is generated and sent on unread.
Brendan spent twelve years running enterprise and international search at Lionbridge, the global localization company, advising Fortune 1000 brands across more than a hundred market and language combinations and reporting findings directly to CMOs.
He built the company's AI search practice from 2023 and wrote the Python tooling behind its multilingual keyword and content work. Before that, he designed data warehouse and business intelligence systems at Dell, EDS/HP and the Revenue Commissioners.
Someone who understands both how the data is structured and how a brand gets found is a rare combination — it's why the technical side of this work holds up.
07 — Free audit
Take a peek at what AI says about you.
Five of your customers' real buying questions, run across ChatGPT, Gemini and Google's AI answers. You get back who's being named, whether you're among them, and an indicative score. No obligation — if there's nothing here worth paying for, I'll tell you that.
TYPICAL TURNAROUND — 3 WORKING DAYS