1. We ask what your customers ask
Every scan starts with a question list built for your market, not a generic template. The questions come in three kinds:
- Discovery. "Who is the best home care agency in Whittier?" The customer has a need and no name in mind.
- Comparison. "Is Harbor Dental better than Bright Smile?" The customer is choosing between options.
- Legitimacy. "Is Harbor Dental a real, trustworthy practice?" The customer is checking before they commit.
Discovery questions never mention your brand. That is deliberate: the only mention that wins a customer is the one the AI volunteers on its own. Each question is asked of ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews through their official APIs, and the exact prompt and full answer are kept so you can read them later.
2. A separate model grades every answer
The engine that answers is never the one that marks its own work. A second model reads each answer and records what it actually contains: whether you are named, which rival brands are named and in what order, every factual claim about your business, and every source the answer cites.
Nothing counts without proof in the answer text. A mention has to be quotable. A wrong fact only counts when the grader can point at the sentence that contains it. This is what keeps the score honest when answers are long and vague.
3. Only earned mentions count
Your score is built from answers to questions that never named you. If an AI names you when asked "tell me about Harbor Dental", that proves it can look you up, not that it would recommend you. Earned mentions are weighted by buying intent, so a recommendation in a discovery answer moves the score more than a passing reference elsewhere.
The same grading produces a share of answers for every brand the engines named, so you see who is winning the questions you are missing.
4. Known, or just findable
Discovery questions run twice: once with live web search on, and once from the model's memory alone. Comparing the two tells you something a single run cannot.
- Named in both: the engine has learned you and can confirm you. This is the goal.
- Named only with web search: you are findable, not yet known. The fix is authority and coverage, so the model absorbs you.
- Named only from memory: the engine remembers you but cannot confirm you live. Usually a crawler or website problem.
- Named in neither: the fix list starts from the beginning.
5. What the AI gets wrong
Every claim the engines make about your business is checked against your own website, with the evidence quoted on both sides. Wrong opening hours, an old address, a service you no longer offer, a misspelled name: each is listed with the page that taught it, so you can correct the source rather than the symptom. Cited pages are also checked for dead links.
6. Your website, audited for AI readiness
Alongside the answers, Renomly reads your site the way an AI crawler would: whether AI bots are allowed in, whether structured data describes the business, whether there is an llms.txt file, and whether the pages answer the questions customers ask in plain language. Backlink authority is checked as well, because engines with web search lean on it when deciding who to trust.
7. The score, and the fix list
Everything above rolls into one number from 0 to 100, with a band that says in words what the number means. The number is useful for tracking; the fix list is what you act on. It is ordered by impact first and effort second, and it starts with your own website because that is the part you control today.
Rescan a topic whenever you want and the score moves in front of you. Plan limits are caps on how much you can run, never overage charges.
What a scan cannot promise
AI answers vary between runs, even for the same question on the same day. A score describes the run it came from, which is why Renomly keeps the full answers and shows the change over time rather than a single snapshot. No tool can guarantee that an engine will recommend you. What Renomly guarantees is that every number on the report has a quote behind it.