Competitor AI visibility analysis
Someone is being recommended. Find out who.
Every day a buyer asks an assistant who the best provider in your category is, and it names three companies. We tell you which three, how often, on which prompts — and the specific assets they hold that you don't.
One named rival · 5 working days · No retainer required
- Competitor A38.2%
- Competitor B26.7%
- Competitor C19.4%
- Industry directory14.1%
- Your brand9.6%
- Competitor D6.3%
The head-to-head
Not “they have more authority”. The exact asset, on the exact prompt.
Pick a rival profile below to see the shape of what we hand over. Yours arrives with real names, real URLs and the passage the model actually lifted.
Why they win
They are the category's default answer, and defaults are self-reinforcing.
An older domain with a decade of trade coverage. Models reach for them because independent sources keep repeating their name — not because their pages are better written than yours. That corroboration advantage is real, and it is also the most attackable thing they have.
- 142 independent mentions across trade press and review platforms
- Cited in 4 of the 6 comparison articles models retrieve most
- Complete Organization + sameAs entity graph, matching knowledge panel
- Glossary hub answering 61 category questions in extractable HTML
Your displacement route
- Their citation share
- 38.2%Their citation share
- Yours today
- 9.6%Yours today
- Prompts winnable in Q1
- 11Prompts winnable in Q1
You will not out-authority them this year and we won't pretend otherwise. You displace them on the 11 prompts where their answer is generic and yours can be specific — pricing, integrations, industry fit and objection queries where a focused challenger beats a broad incumbent.
priority: comparison + objection prompts · effort: medium · projected Δ +14–19pts by day 90
Run this against my real rivalWhy they win
Newer, smaller, and beating you because their pages are built for retrieval.
Lower domain authority than yours and a fraction of the backlinks — yet cited twice as often. Their content answers in the first forty words, one claim per block, in extractable HTML. Yours buries the same answer under three paragraphs of preamble and a JS accordion.
- Answer-first structure on 34 of 40 commercial pages
- Zero render dependency — full content in the initial HTML payload
- FAQ and HowTo schema validated across the whole template
- dateModified maintained on every page, so temporal prompts favour them
Your displacement route
- Their citation share
- 21.8%Their citation share
- Yours today
- 9.6%Yours today
- New links required
- 0New links required
This is the cheapest gap on the board because it is entirely structural. Same claims, same authority, different packaging. We rewrite your top pages passage-first and fix the render path — no new links, no PR budget, no waiting on domain authority to compound.
priority: structural rewrite + render fix · effort: low · projected Δ +12–16pts by day 60
Run this against my real rivalWhy they win
A directory is holding the slot, and it is beatable.
Half your category's citations are going to a listicle that does not sell anything. Models like it because it is comprehensive, current and unbiased-looking. The good news: aggregators are shallow. They cannot answer a specific buyer question, and they never get named alone.
- "Top 20 providers" pages refreshed monthly with visible dates
- Comparison tables in plain HTML — trivially extractable
- Neutral framing models treat as low-risk to cite
- Broad coverage of long-tail category and location prompts
Your displacement route
- Directory share
- 14.1%Directory share
- Yours today
- 9.6%Yours today
- Long-tail prompts open
- 23Long-tail prompts open
Two moves. Get listed and ranked inside the aggregator itself so their citation carries you, then take the 23 specific prompts they cannot answer — implementation, pricing model, industry fit — with depth a listicle structurally cannot match.
priority: listing placement + long-tail depth · effort: low-medium · projected Δ +9–13pts by day 90
Run this against my real rivalMethod
Six models, sampled — not one screenshot of one chat
- 01
Competitor set definition
Your named rivals plus the ones actually appearing in answers — which is rarely the same list. Aggregators and adjacent categories included.
5–12 competitors · client + discovered - 02
Shared prompt set
One set of 400+ commercial prompts run identically for every brand, so the comparison is genuinely like-for-like rather than cherry-picked.
400+ prompts · identical across brands - 03
Multi-model sampling
Every prompt run n≥30 per model at controlled temperature and locale. Share of voice is reported as a rate with a confidence interval.
6 models · n≥30 · ±4.1% at n=30 - 04
Asset differential
For every prompt they win, we identify the cited URL, the retrieved passage, and the structural or authority advantage that produced it.
URL + passage level attribution - 05
Corroboration mapping
Which third-party sources mention them and not you, ranked by how often those sources appear in your category's citation sets.
source graph · frequency weighted - 06
Displacement modelling
Every winnable prompt scored on effort against projected share gain, so the plan is ordered by return rather than by what's easiest.
ICE-scored · projected Δ share
What lands in your inbox
Six artefacts your team can act on the same week
- Dashboard
Category share-of-voice board
Every competitor's citation share by model and prompt cluster, with your position and the gap to each rival.
For your CMO - CSV + JSON
Prompt-level win/loss matrix
Every prompt, who was cited, in what order, across all six models and every sampled run.
For your analytics team - Head-to-head
Asset differential per rival
What each competitor holds that you don't — pages, schema, mentions, structure — quoted with URLs.
For your strategist - Ranked plan
Displacement roadmap
The winnable prompts in priority order, each with the specific asset required and the effort estimate.
For your head of marketing - Source graph
Corroboration gap list
The third-party properties citing them and not you, ranked by influence on your category's answers.
For your PR lead - 60 minutes
Analyst readout
A working session with the analyst who ran it — not an account manager reading a slide.
For everyone above
Proof
Three categories where the leader changed
- B2B SaaS
Lumen
Losing every comparison prompt to a rival with half their traffic. The differential showed it was structure, not authority — 34 answer-first pages against their 6. Rewrote 22 pages; no new links.
9% → 41%Citation share5 months - Professional services
Northbay
A directory held 31% of category citations. We got them listed and ranked inside it, then took 23 specific prompts the listicle structurally couldn't answer.
+27ptsShare of voice6 months - Manufacturing
Atlas+Co
Their named rival wasn't the one models cited. The real competitor was an adjacent-category player nobody in the business had heard of. Strategy changed in week two.
1stCategory rank4 months
Investment
Benchmark once, or track the category continuously
The head-to-head is complimentary and arrives before any contract. Everything beyond it is quoted after we have seen the size of your competitive set.
- Start here
Head-to-Head
One named rival, benchmarked against you across every answer surface.
Complimentary5 working days- 1 competitor, 5 commercial prompts
- All 6 models sampled
- Both citation shares side by side
- Asset differential summary
- Top 3 winnable prompts
- 30-minute findings call
- Fixed scopeMost chosen
Category Analysis
Your whole competitive set measured once, with a displacement roadmap you can hand to any team.
By quoteone-off · 10 business days- Up to 12 competitors
- 400+ prompts, n≥30 per model
- Full win/loss matrix (CSV + JSON)
- Asset differential per rival
- Corroboration gap list
- 60-minute analyst readout
- Ongoing
Category Watch
Continuous tracking plus the displacement work, re-measured every month.
By quotemonthly · 90-day term- Everything in Category Analysis
- Monthly re-run on the same prompt set
- Competitor movement alerts
- Content and schema work delivered
- Corroboration and digital PR outreach
- Monthly analyst review call
Why us
Four reasons clients bring this to us
- 01
You see the gap before you commit
The head-to-head lands before any contract. If your category isn't being answered by assistants yet, we tell you and stop there.
- 02
Sampled, not screenshotted
Every number is a rate with a confidence interval across thirty runs per model — not one lucky chat session in a slide.
- 03
In-house, not brokered
Analysts, content, technical SEO and digital PR sit in one team. Nothing is subcontracted and nothing waits on a third party.
- 04
Yours to take anywhere
The analysis is a deliverable, not a hostage. Hand it to your own team or another agency — plenty of clients do.
“We assumed our biggest rival was the one we'd been benchmarking against for years. The analysis showed the brand taking our slots was one nobody in the room could name.”
Before you ask
The questions that decide it
Answer
How is this different from a normal competitor audit?
A normal audit compares rankings, backlinks and traffic estimates — inputs. This measures the output that now decides shortlists: how often each brand is actually named and cited when a buyer asks. Those two rankings often disagree sharply. We routinely find the domain-authority leader sitting fourth in citation share, beaten by a smaller rival with better-structured pages.
Ask us something harderAnswer
What if we don't know who our real competitors are?
Most clients don't, and that's one of the more useful findings. We build the set from your named rivals plus everyone actually appearing in your category's answers — which regularly surfaces an adjacent-category player or an aggregator nobody internally was tracking. In about a third of engagements the discovered competitor matters more than the named one.
Ask us something harderAnswer
Can you tell us exactly why a competitor was cited?
We can't read the model's internals and we don't claim to. What we do is embed the candidate passages, compare each against the prompt, and cross-reference the cited URL and quoted fragment in the response. That yields a defensible best-candidate attribution with a stated confidence — enough to act on, because the remediation is identical whether confidence is 0.8 or 0.95. We report the number rather than hiding it.
Ask us something harderAnswer
Isn't citation share a vanity metric?
It is if you stop there. We tie it to prompt intent, so you can see share on high-intent commercial prompts separately from informational ones — being cited on 'what is X' is worth far less than being cited on 'best X for Y'. Then we map AI-referred sessions and pipeline in GA4. Share of voice is the leading indicator; enquiries are the number you judge us on.
Ask us something harderAnswer
Do you also do the work, or just the analysis?
Both, and you can take either. The analysis is a fixed-scope deliverable you're free to hand to your own team or another agency — plenty of clients do exactly that and we're fine with it. If you'd rather we execute, the Category Watch and displacement programmes include the content, schema and corroboration work with monthly re-measurement.
Ask us something harderAnswer
How often does this actually change?
Meaningfully, quarter to quarter. Model updates, competitor publishing and new third-party sources all move it, and we've seen a category leader lose eight points of share in six weeks after a rival shipped one well-structured comparison hub. A one-off analysis is a snapshot with a shelf life of roughly a quarter — which is precisely why the ongoing programmes exist, and also why we'll tell you if you only need the snapshot.
Ask us something harderName one competitor. We'll tell you why they're winning.
Send your domain, one named rival and five commercial prompts. You get both citation shares side by side, the assets they hold that you don't, and the three prompts you could take first. Five working days, no retainer required.
3 brands named in a typical AI answer. That's the entire shortlist · 71% of buyers who ask an assistant never open a second source · 1 in 3 engagements find a competitor the client wasn't tracking
Works hardest alongside
All AI search services- Parent hubAI Search VisibilityThe full programme across every AI answer surface.
- Sibling serviceLLM Citation AnalysisSource-level auditing of which URLs models actually cite.
- Sibling serviceGoogle AI OverviewGet cited in the answer box above your own rankings.
- Sibling serviceChatGPT VisibilityBe named when buyers ask an assistant instead of Google.
- ProgrammeAI SEORankings and retrieval run as one engagement.
- AuthorityDigital PRThe third-party corroboration models weight most.
