How your agency shows up in AI search.
Your next brief starts with a question typed into an assistant. It answers with 3 agency names. This is how you find out whether you are one of them, and what decides it.
Book the £999 audit → See the 5 signals →The shortlist is now a paragraph.
For 20 years agency business development meant a phone, a list, and a reason to call. That still works. What has changed is what happens before the call, and increasingly instead of it.
A hiring manager with a vacancy and no incumbent supplier asks an assistant. They get one answer with 3 names in it. There is no page 2. There is no shortlist of 10 links to work through. You are in the paragraph or you are not in the process.
- A competitor
- Another competitor
- A third you have never lost to
- Your agency : not cited
Illustrative example, not a live reading.
Your competitors are not outbidding you. They are being recommended, and you are not being mentioned.
Clients ask harder questions than candidates do.
Almost everything written about AI visibility is aimed at employers worried about candidates. That is a real problem and we cover it separately in employer brand AI visibility.
The agency version has a different buyer and a higher bar. An employer wants candidates to find them. You want clients to find you, and a client is not asking what it is like to work somewhere. They are asking who fills roles like mine, in my sector, in my region, at my volume, and can that be checked.
An assistant answering that is hunting for evidence of specialism. Most agency websites publish evidence of existence.
That is the whole gap. Sector pages that could belong to any agency. A case study with the client name removed and the number rounded to nothing. An about page written for investors. None of it gives a model anything it can safely repeat with your name attached.
5 things decide whether an assistant can name you.
They compound, and they are ordered here by dependency rather than weight. Miss the first and the rest are academic, because the model never reaches you.
Access
Whether the AI crawlers can reach your content at all, or your robots file is quietly turning them away.
Agencies hide their best proof behind JavaScript job boards and gated PDFs.
Structure
Whether your content is marked up so a machine understands what it is reading.
Most agency sites mark up JobPosting and nothing else. Some not even that.
Knowledge graph
Whether a clean, claimed entity for your business exists where models go for facts.
Trading name against Companies House name is the commonest agency break.
Citation
Whether anyone other than you says you are good at what you say you are good at.
Where agencies score worst, and where the score is least reliable. See below.
Social
Whether the third party signals a model ingests say your business is real and current.
A dormant company page reads as a dormant company.
These are the 5 dimensions The SetpointHQ Index scores. In the published rubric, citation carries the heaviest weighting at 35%, schema 20%, and the remaining 3 at 15% each. Read the methodology →
Your citation score is a floor, not a reading.
The Index scores 100 UK employer brands. Agencies are not in that cohort, so there is no sector median for you to sit against and we will not invent one. You get the score, the evidence and the fix list. You do not get a cohort rank, because it would not mean anything.
The citation probes are also written the way a candidate asks about an employer. Run against an agency they under count, because the questions clients ask are not the questions the rubric asks. An agency citation score is a lower bound. Anyone selling you an agency AI visibility score without saying that is selling you a number they cannot stand behind.
Employer cohort, rubric 2026.06.5, 13 July 2026. Source: The SetpointHQ Index. Agencies are not scored in this cohort.
5 failures we see on nearly every agency site.
The job board eats the site
Live vacancies render in JavaScript from an ATS on a subdomain. To a crawler the careers content is a blank div, and the vacancies are the only page on the site that proves what you actually recruit.
The name does not match anywhere
You trade as one thing, you are registered as another, LinkedIn says a third, and your invoices say a fourth. A model resolving an entity treats that as 4 weak candidates rather than 1 strong one.
The proof is in a PDF
Your best case study is a gated download. The page that would earn the citation is a form. The evidence exists and no model has ever read it.
Sector pages that say nothing
Swap the sector noun and the page still reads fine. If a page works for warehousing and social care without edits, it is not evidence of specialism in either.
No numbers anywhere
Fill rates, time to hire, volumes placed, retention at 12 months. Models repeat specifics and skip adjectives. Every claim on your site that a competitor could also make is a claim that will never be quoted.
3 moves, in this order.
1. Find out where you stand
Ask the 4 assistants the questions your clients ask, using your sector and your region, and write down whether you appear. It takes an afternoon and it is the only part nobody can do for you. If you are missing, start with the 6 reasons an agency goes missing from ChatGPT.
2. Fix the machine readable layer
Unblock the crawlers. Get the entity clean and consistent across Companies House, LinkedIn and your own site. Mark up the organisation, not just the jobs. This is a fortnight of unglamorous work and it is the part that gates everything else.
3. Publish evidence a model can quote
Ungated case studies with real numbers. Sector pages that would break if you swapped the sector. Named results where the client will let you name them. This is the slow part, and it is the only part your competitors cannot replicate in a fortnight.
There are 20 years of recruitment marketing behind this, spent working for recruiters rather than as one. That includes a social care agency taken from £12m to £100m+ over 12 years. The pattern held then and it holds now. The firms that win the brief are the ones whose proof is easiest to check.
Four parts of the picture.
Your agency is not showing in ChatGPT
6 reasons, in the order to check them, with what each one looks like when it is the culprit.
Read →
Generative engine optimisation for recruitment
The discipline that replaces SEO when the buyer gets one answer instead of 10 links.
Read →
Employer brand AI visibility
The client side of the same problem, for in house teams worried about candidates rather than briefs.
Read →
Is your company visible in ChatGPT?
A 5 minute self check you can run today, with the exact prompts to use.
Read →
Common questions.
How is agency AI visibility different from employer brand AI visibility?
The mechanics are the same 5 signals. The buyer is not. An employer is trying to be recommended to candidates, so the questions are about pay, culture and progression. An agency is trying to be recommended to clients, so the questions are about sector, region, volume and provable results. The evidence a model needs is different, and most agency sites publish neither set.
Can we pay to be recommended by an assistant?
No. There is no sponsored slot inside a synthesised answer. The model weighs what it can reach, parse and corroborate about you. That is the whole reason this is worth doing early, because the advantage compounds and cannot be bought back quickly.
Which assistants should an agency care about?
ChatGPT, Claude, Perplexity and Google Gemini, plus the AI Overviews sitting on top of Google search. Those are the 4 The SetpointHQ Index probes.
We rank well in Google. Is that not the same thing?
No. Ranking ninth still put you on the page. Being named ninth in a 3 name answer does not exist. Search rewarded presence and you could buy position. Assistants reward corroborated evidence and you have to earn it.
How long before this shows up in briefs?
It already does, and it is invisible in your data. A client who never contacts you because an assistant did not name you leaves no trace in your CRM. That is the difficult part of measuring it, and the reason most agencies have not noticed yet.
Does the £999 audit cover agencies, or only employers?
It covers agencies. The measurement is the same: all 5 dimensions scored across the 4 assistants, the evidence behind each number, and a prioritised fix list, back in 7 working days with a 60 minute debrief. You do not get the cohort comparison, because The Index cohort is 100 employer brands and ranking you inside it would be meaningless. The questions we probe with are yours, built from the briefs you want to win rather than the candidate questions the employer rubric uses.
Find out whether the assistants name you.
The £999 Recruitment Marketing Audit scores all 5 dimensions across ChatGPT, Claude, Perplexity and Gemini, shows you the evidence behind every number, and hands back a prioritised fix list in 7 working days with a 60 minute debrief. Probed with the questions your clients ask, not the ones a candidate would. If the honest answer is that you have bigger problems than this, we will say so on the call.
Book the £999 audit → Explore Studio, £149 per seat →