Your customer asked ChatGPT. It named three competitors.
People no longer scroll through search results — they ask an assistant for a recommendation and act on the first name it gives. We measure what ChatGPT, Perplexity and Gemini say about your category, then fix what stops them saying your name.
No retainer to start. The audit is free because it is the fastest way to show you the gap — and you keep it whether you hire us or not.
An answer engine names three businesses. Everyone else is invisible.
Ask for "who does SR-22 filings in Long Beach" and you do not get ten links. You get a paragraph naming a few businesses the model can identify with confidence. There is no page two, and no ad slot to buy.
The model must resolve you
It has to know your website, your Google Business Profile and your social pages describe one real business in one place. Contradictions make it skip you.
The model must be able to quote you
Answers are assembled from content that already states facts plainly. Brochure copy that never says what you do, where, and for whom is unusable.
The model must have corroboration
Reviews, listings and third-party mentions are the evidence it weighs. A claim with nothing behind it reads as a marketing statement, not a fact.
What actually decides whether you get named
Five signals, and most businesses fail at least three of them. Every audit checks all five.
| Signal | What it means | Common failure |
|---|---|---|
| Entity resolution | Website, Google Business Profile, Facebook and Yelp reference each other as the same business | No sameAs, conflicting addresses or phone numbers |
| Structured data | Machine-readable description of services, area served, hours, coordinates | No schema, or areaServed naming a whole state |
| Answer-shaped content | Pages that state the answer in the first 40–60 words, then explain | Marketing pages with no direct answer to a real question |
| Corroboration | Reviews, citations and mentions that confirm the claim elsewhere | No reviews published, no listings claiming the business |
| Crawler access | AI crawlers can read the page as HTML, not an empty JavaScript shell | Client-rendered sites that serve bots a blank page |
Three steps, and you see the evidence before you pay
We ask what your customers ask
Twenty fixed prompts in your category — English and Spanish where relevant — run against ChatGPT, Perplexity and Gemini. The same twenty every month, so the comparison is real.
You get the recording and the ranking
A ten-minute walkthrough: were you named, at what position, described how, and precisely which competitors took the slots that should have been yours.
We fix the machine, not the copy
Entity consolidation, structured data, answer-shaped content, review engine, crawler access. Then we re-run the same prompts and show you what moved.
Who this is for
Businesses where one customer is worth real money, and where people ask an assistant before they call.
- Insurance agencies — SR-22, non-owner, high-risk, auto and DMV services
- Personal injury and immigration law — high-intent, heavily compared categories
- Dental, med-spa and cosmetic clinics — recommendation-driven, review-sensitive
- Roofing, HVAC and home services — local, urgent, price-comparable
- B2B software — where buyers ask an assistant to shortlist vendors
The gap is almost always the same
Ask a model about the business by name and it usually knows them — 4 out of 4 prompts answered. Ask it for the best option in the category and they vanish.
| Brand prompts answered | 4 / 4 |
| Category prompts answered | 0 – 2 / 16 |
The model knows exactly who they are and still recommends somebody else. That gap is the whole problem, and it is measurable.
Pricing
Start with the free audit. Most clients then take the sprint, and a minority keep the monthly tracking.
Visibility audit
20 prompts × 3 models, presence scoring, competitor leaderboard, entity diagnostic, prioritised fix list and the recorded walkthrough.
Request the auditImplementation sprint
Entity and schema rebuild, llms.txt, answer-structured content, Google Business Profile, review engine, English/Spanish alignment. Three to four weeks.
Tracking & content
Monthly re-run of the same 20 prompts, citation tracking, competitor movement, and four answer-first pages a month.
See detailsQuestions worth asking
What does "recommended by an LLM" mean?
When someone asks an assistant for a recommendation in your category, the answer names a handful of businesses. Being named is being recommended. Most businesses are named for their own brand and never for the category.
Can you guarantee we'll be recommended?
No. Answers change weekly and vary by phrasing. Anyone selling a guaranteed position is either measuring badly or lying to you. We commit to measurable movement across a fixed prompt set.
How is this different from SEO?
SEO competes for a rank in a list of links. This competes for a sentence. The work overlaps — schema, entity consistency, content — but a page-one ranking does not put your name in an AI answer.
Which models do you test?
ChatGPT, Perplexity and Gemini, with browsing or grounding enabled, version-stamped and dated so month-on-month comparisons stay valid.
How long before something changes?
Technical and entity fixes are indexed in days and can affect retrieval within weeks. Model answers move more slowly. We report monthly and don't promise dates.
Do you work in Spanish?
Yes — audits include Spanish prompts and we deliver Spanish content. In many US markets almost no competitor publishes answer content in Spanish, which makes it the cheapest advantage available.
Find out what the AI says about you
Twenty prompts. Three models. A recording, a competitor list and the exact fixes. Free.
Request the free auditOr email hello@llmrecommendations.com