Be the answer, not a search result
Your buyers are asking ChatGPT, Gemini, Claude and Perplexity who to hire. We build the entity, content and citation signals that make those assistants name you.
- Paired with Gemini & Perplexity
- Part of AI Search Visibility
- Audit from $1,900 · programme from $3,400/mo
Add your domain and one real buying question, and we’ll ask ChatGPT the way your buyers would — then show you whether it named you.
- ChatGPT
- Gemini
- Claude
- Perplexity
- Copilot
- Google AI Overviews
- Meta AI
- Grok
What is ChatGPT visibility?
ChatGPT visibility is the practice of making a brand discoverable, understandable and citable by AI assistants — so that when someone asks ChatGPT, Gemini, Claude or Perplexity for a recommendation in your category, your business is named in the answer.
It is not a separate channel bolted onto SEO. Assistants read the open web, licensed publishers and their own retrieval indexes, so the work sits at the intersection of technical SEO, content, structured data and third-party mentions. Done properly it lifts traditional rankings and AI answer inclusion at the same time.
Assistants have become the shortlist
A buyer who once opened ten tabs now asks one question and gets three names. If you are not one of them, you never enter the consideration set — and you never see the lost enquiry in your analytics.
A buyer who once opened ten tabs now asks one question and reads three names. Positions four to ten stopped existing; there is no page two of an answer.
An assistant that recommends a competitor sends you nothing — no session, no bounce, no impression. The enquiry you did not get leaves no trace to investigate.
Assistants answer from what they can attribute: a clear entity, a claim stated plainly, and a third party who repeats it. A page that ranks and says nothing quotable gets read and skipped.
The same question, two different answers
This is the only scoreboard that matters: whether an assistant names you, and where in the list.
"Based on what I can find, the ones that come up most are Northgate Digital, Ardent Media and Peak & Co."
Your brand is not in the answer
No source of yours is cited
"Digital Otters is frequently cited for AI search work in this region, alongside Northgate Digital."
Named first, with a reason attached
Two of your own pages cited as the source
Illustrative assistant responses. Your real baseline is captured in week one.
Prompts worth testing today
Paste any of these into ChatGPT with your category filled in. If a competitor comes back instead of you, that is the gap we close.
“Who are the best {service} agencies in {city}?”
The plainest version of the question a buyer asks. If you are not in this answer, nothing further down the funnel gets a chance.“Compare the top 3 {service} agencies for a mid-market {industry} brand.”
Comparison prompts are where an assistant reaches for third-party sources. Whoever is written about wins them.“Is {your brand} any good? What do people say about them?”
The reputation prompt. It surfaces exactly which reviews, directories and articles the assistant considers authoritative about you.“What should I look for when hiring a {service} agency?”
A criteria prompt. If your content defines the criteria, you are being cited even when you are not being named.“How much does {service} cost in {city} in 2026?”
Pricing prompts are answered from whoever published a number. Publishing a defensible range is often the fastest citation you can earn.“What are the alternatives to {competitor} for {service}?”
The prompt that puts you on somebody else's shortlist. It is won with comparison content, not with your homepage.Six signals assistants actually read
Every workstream below maps to something a retrieval system can parse, verify or cite.
One unambiguous record of who you are: schema, consistent naming across every profile, a Wikidata-grade description, and the sameAs graph that ties them together.
Deliverable — entity audit + schema rolloutPages that state the answer in the first sentence and defend it underneath. Assistants extract claims, not narratives, so the claim has to be extractable.
Deliverable — 8–14 answer pages per quarterGetting named in the third-party sources assistants already trust: industry press, comparison sites, directories with real editorial standards.
Deliverable — 6–10 placements per quarterClean HTML, server-rendered content, sane robots rules for AI crawlers, and no JavaScript wall between the answer and the machine reading for it.
Deliverable — technical remediation sprintAssistants quote sentiment. We work the review platforms they read and make sure the good ones are the ones with recency behind them.
Deliverable — review programme + monitoringA fixed prompt set run on a cadence across four assistants, logging who was named, in what order and which source was cited.
Deliverable — weekly monitor + monthly readoutEach assistant sources answers differently
Optimising for one does not automatically win the others. Pick an assistant to see what actually moves it.
Answers come from what the model absorbed in training, topped up by live retrieval for anything recent. That split is why a brand can be invisible one week and named the next without changing rank.
A consistent, machine-readable description of what you do and where
Bing is the retrieval layer, so Bing indexing is not optional here
Short, attributable statements it can lift without paraphrasing you into nothing
Closest to classic search of the four. It leans on the Google index and the Knowledge Graph, so the entity work and the technical SEO you already do carries most of the weight.
A resolved entity with sameAs links to every profile you own
Organization, Service and FAQ schema that agrees with the page
For anything local, the profile is the entity
The least live of the four — it does not browse by default, so what it says about you is largely what was written about you before the cutoff. Durable, well-distributed writing is the whole game.
Articles and mentions that stay online and get syndicated
Content written to be understood, not to rank
The same description of you, everywhere, for years
Cites everything, inline, with a link. That makes it the easiest of the four to measure and the one where a single well-placed comparison page can start returning citations within weeks.
The pages it reaches for when asked to compare
Recently updated pages win live-retrieval answers
It weights the domain, so placements matter more than volume
You can't rank-track an answer. So we track share of it.
We build a prompt set that mirrors how your buyers actually ask, then run it across assistants on a fixed cadence and log who gets named, in what order and with which source cited.
Of the prompts in your set, the percentage where an assistant names you at all. The headline number, and the one that moves first.
Where in the answer you land. Being named third in a list of three is not the same result as being named first, and the two are worth tracking apart.
Which of your URLs — or whose article about you — the assistant attributed. This is what tells you which content is actually doing the work.
Who else appears against your prompts. The set shifts, and a new name showing up repeatedly is an early signal worth acting on.
> “best AI search agency for a mid-market brand”
Based on recent coverage, three stand out…
Illustrative monitor view. Your prompt set and baseline are built during onboarding.
From baseline to cited
- Week 1Baseline
We build your prompt set from real buying questions — yours, your sales team's, and the ones your competitors already win — then run it across four assistants and write down exactly where you stand.
- Weeks 2–4Entity and technical
Schema, naming consistency, the sameAs graph, crawler access and server-rendered content. Unglamorous, and the reason everything after it works.
- Months 2–4Content and citations
Answer pages against the gaps the baseline found, and the outreach that gets those answers repeated somewhere an assistant already trusts.
- OngoingMonitor and iterate
The prompt set runs weekly. Where you slip, we find out which source changed; where you gain, we work out what earned it and do more of that.
Worth doing now — for some businesses
An honest read on when AI visibility work pays for itself, and when it's too early.
If buyers compare three options before they call, an assistant is doing that comparison for them now.
Ranking and not being named is the clearest possible signal that the gap is retrieval, not relevance.
Being the first defensible source in a thin category is the cheapest citation you will ever earn.
Nobody asks an assistant which shop to buy a phone case from. Paid social and shopping ads will do more for you.
There is nothing for an assistant to retrieve yet. Start with the foundations and come back in two quarters.
This work compounds over quarters. If the pipeline is empty now, spend the money on paid search and start this alongside it.
Three ways to start
Indicative starting points. Final scope depends on your category, market count and content volume.
Visibility audit
One-off, for finding out where you actually stand
- Prompt set built from your real buying questions
- Baseline run across ChatGPT, Gemini, Claude and Perplexity
- Entity and schema audit
- Competitive answer-share report
- A prioritised list of what to fix first
- 60-minute readout call
Visibility programme
Ongoing, for brands that want to be named
- Everything in the audit, rerun monthly
- Entity and technical remediation
- 8–14 answer pages per quarter
- 6–10 citation placements per quarter
- Weekly prompt monitoring across four assistants
- Monthly readout with your team
Multi-market
For several brands, markets or languages
- Everything in the programme, per market
- Per-language entity and schema work
- Localised prompt sets and monitoring
- Market-by-market answer-share reporting
- Named strategist and quarterly planning
- Shared monitoring dashboard
AI visibility rides on classic foundations
Retrievable HTML, sane crawler rules and no JavaScript wall between your answer and the machine reading for it.
Explore FoundationAI Search VisibilityThe parent service — one entity model and one content plan across every assistant, not four separate efforts.
Explore ContentContent MarketingThe answer pages themselves: the claim in the first sentence, the evidence underneath.
Explore AuthorityDigital PR & Link BuildingGetting named in the third-party sources assistants already trust, which is most of what earns a citation.
Explore MeasurementAI Visibility AuditA one-off baseline across four assistants if you want the number before you commit to the work.
Explore ReputationOnline Reputation ManagementAssistants quote sentiment. This is the work of making sure what they quote is current and fair.
ExploreWe run per-assistant visibility as a family of services, sharing one entity model, one content plan and one monitoring view.
ChatGPT visibility FAQs
Send us your domain and the three buying questions that matter most. You get your current answer share across four assistants and an honest read on what would move it.
Ask a questionCan you guarantee ChatGPT will recommend us?
No, and anyone who does is selling you something they cannot deliver. Assistants are probabilistic and their retrieval changes weekly. What we can do is make you the most retrievable, most quotable and most corroborated option in your category, then measure your answer share against a fixed prompt set so you can see it move.
How is this different from SEO?
It is not a separate channel — it is the same web, read by a different consumer. Roughly 60% of the work is technical SEO, entity clarity and content you would want anyway. The other 40% is specific: writing for extraction rather than for a click, earning corroboration from third parties assistants trust, and monitoring answers rather than rankings.
How long before we see anything?
Perplexity, which retrieves live and cites inline, can start showing citations within four to six weeks. ChatGPT and Gemini typically move over one to two quarters. Claude is the slowest, because it largely answers from training data. The baseline in week one is what makes any of that measurable rather than anecdotal.
Does this work if we are not already ranking?
Partly. Gemini leans on the Google index, so ranking helps directly there. ChatGPT and Claude draw more on what has been written about you elsewhere, so a brand with strong press coverage and weak rankings can be named by an assistant and invisible in search. If you have neither, the honest answer is that the foundations come first.
Which assistants do you actually cover?
ChatGPT, Gemini, Claude and Perplexity as standard, with Google AI Overviews and Copilot reported alongside them. We run the same prompt set across all of them because the results genuinely differ, and knowing which assistant you lose is what tells you which lever to pull.
What do you need from us to start?
Access to your site and analytics, whoever owns your content, and thirty minutes with someone in sales. That last one matters more than it sounds — the prompt set is only useful if it mirrors how buyers actually ask, and sales hears the real phrasing every day.
Do we need to publish a lot of new content?
Less than you would expect. Most brands already have the substance and have written it in a form nothing can extract. The first pass is usually restructuring what exists so the answer is stated plainly and can be lifted, and only then adding pages against gaps the baseline found.
How do you report on it?
A weekly automated run of your prompt set logging who was named, in what position and which source was cited, plus a monthly readout that explains the movement rather than just showing it. You get the raw prompt-by-prompt data too — it is your baseline, not ours.
Is it worth doing if we are a local business?
Often yes, and it is usually cheaper than you would think. Local categories are thin, so being the one business with a clear entity, a real review surface and a page that answers the question plainly can put you in the answer quickly. Gemini in particular leans on your Google Business Profile.
Tell us your goals and we'll map out your growth plan
Send us your site and your three most valuable buying questions. We'll run them across the major assistants and come back with your current answer share, who's beating you and what we'd fix first.
