Build the map question by question, not practice by practice. For every question where a model recommends somebody else, find the source it used: a directory profile, a review page, or one treatment page the competitor owns and you do not. That source is your work list. A map that only says which practices outrank you is a mood, not a plan.
Competitors Are Not a Market Share Problem
A practice can be missing from most of the answers in its city and still be in no trouble at all. What matters is whether the missing answers are the ones your patients act on.
answers name your practice
answers name your practice
Patients do not ask for a list of every dentist in the area. They ask who can see me today, who does implants well, who takes my insurance. Each of those is a separate contest with a separate winner. Losing the insurance answer while winning the same week appointment answer is not a crisis. Losing both is.
That changes the accounting. Instead of asking whether you are ahead of [Competitor A], you ask which of your twenty questions [Competitor A] won, and why. The competitor is not the subject of the exercise. The question is.
The Two Piles Of Questions
Sort your twenty questions into two piles: the ones where a reader would book this week, and the ones where they would bookmark for later. The first pile is worth more per win and should be fixed first, even when a later question is easier to fix. A practice that wins every comparison question and loses every emergency question is invisible to the patient in pain.
Step 1: Find The Competitor Set That Actually Wins
Most practices build a competitor list from rivalry rather than from data: the practice they dislike, the one across the street, the one running the adverts. Build it from the answers instead.
Where The List Comes From
- Run your question set in your city, three runs per engine.
- Write down every practice name that appears in an answer where you do not appear.
- Count the appearances and keep the names that come up on more than one question.
- Discard names that appear only inside a generic list with no reason attached. They are being listed, not chosen.
- Keep three to five practices. More than that and a map stops being a map.
The list you end up with rarely matches the list you expected, and that is the point. Our measurement guide covers how to capture the runs that feed this step in the first place.
Step 2: Map The Win To a Source
For every lost question, read the answer again and find what it pointed at. Answers usually cite something specific: a profile, a directory, a service page, a review thread. Write the source next to the question, because the source is the fix.
Three Sources Do Most Of The Work
- Directory and map listings. Engines reach for these when the question is local and the answer is a place rather than a procedure. If a competitor's listing carries detail yours does not, you have a facts problem rather than a content problem.
- Review profiles. Review volume and recency carry the trust signal, and the trust signal is what turns a name into a recommendation. How directory vitals feed AI answers covers the mechanics.
- Treatment pages. A dedicated page for one procedure is the most quotable thing a practice can own, and it is the one most often missing.
Worked example. You asked which dentists in [CITY] do the most implant work. Gemini returned three practices, one of them a competitor, and gave the reason: that practice had a dedicated implant page with the procedure described step by step, and a review profile carrying a large volume of implant reviews. You were not mentioned at all. The diagnosis is not that the competitor has better marketing. It is two plain facts: you have no implant page, and you have no implant specific reviews under your name. The first is a publishing job. The second is a conversation you are probably already having after every implant case.
Do that for every lost question and the work list writes itself, because each entry now has a named cause. For the page side of it, see how topical authority is built for dental topics.
Step 3: Classify The Gap
Every gap falls into one of five types, and each type has a different fix. Diagnose the type before you spend anything.
Swipe the table sideways to see every column.
| Gap Type | Symptom In The Answer | What Actually Fixes It |
|---|---|---|
| No page | Your service is never mentioned, and a competitor's treatment page is cited instead | Publish one page for that treatment: what it involves, who performs it, what it costs you to have done |
| Thin page | Your page exists, but the answer still names the competitor | Rewrite around the questions a patient asks, with specifics an engine can lift |
| Fact conflict | The answer contradicts your details: wrong hours, wrong services, a second location | Correct whichever source carries the wrong fact, starting with the directory profiles |
| Missing review proof | The competitor is named for trust reasons and you are not | Ask for a specific review after a specific visit, and stay inside what the FTC Rule On Reviews and Testimonials allows |
| Wrong entity data | The model cannot tell who you are, or merges you with a similarly named practice | Fix the entity markup and the name consistency first, then re-test the question |
The row order is not important, the diagnosis is. A gap that looks like thin content is often a fact conflict, and a page rewritten twice will not fix a practice the engine cannot identify. Check the entity row first, every time.
Mark a treatment page up as a medical procedure so it describes itself in a form a machine can read, keep the business profile details identical to the page, and put the review request into the routine you already have rather than treating it as a campaign. Patient engagement tooling such as Engage OS is built for that part of the work.
Step 4: Turn The Map Into a Work List
Order the work by two keys: questions lost, and effort. Highest traffic intent first, because an emergency answer is worth more than a comparison answer even when it is harder to fix.
The Order Of Operations
- Emergency questions. Fewest of them, worth the most, and usually decided by facts rather than prose.
- Treatment questions. One page per treatment, written to be quotable rather than to be read.
- Price questions. Say what a visit includes and what varies. An honest range, clearly marked as an estimate, is more quotable than a number nobody will publish.
- Insurance and availability questions. Usually a facts fix, usually quick, and they vanish the moment your directory and your site disagree.
- Open and neighbourhood questions. Cheapest to skip, most expensive to ignore, and the one place you can be named where the model names nobody.
For anything touching your own treatment or what it will cost you, ask your own dentist; this list is about how people search, not about clinical advice.
Then put the work list where one named person owns it, with a date against each line, and re-run the same questions when it is done. The measurement method shows how to capture a clean before and after, so the next decision rests on data rather than on memory.
How Often To Rebuild The Map
Monthly for the top five competitors, quarterly for the full set. Monthly because a competitor can win a question by publishing one page, and that happens inside a month. Quarterly for the full set because the tail rarely moves, and rescoring forty practices on a fixed schedule costs a day you could spend fixing.
When a Change Is Real
Resist rebuilding everything each time the map shifts. A new practice appearing on three or more money questions deserves a full diagnosis. One new appearance on one question is a data point: note it and check again next month. Rebuilding the whole map for every wobble produces a document nobody reads and a team that stops trusting it.
If you would rather have the map built for you, the full audit scores the competitors taking your answers alongside your own practice, and the case studies show how gaps like these were closed.