You do not need a paid AI answer tracking tool to know what ChatGPT, Gemini, Perplexity and Claude say about your dental practice. One spreadsheet, one fixed run window and a saved copy of every answer gives you the same evidence, and the whole monthly routine takes about an hour.
Most tools in this category sell a score with a logo on it. Scoring is the cheap part. The hard part is deciding in advance what to record so that, three months later, you can prove something changed.
Weekly Spot-checks That Catch Changes
A tracker records answers, not opinions about answers. The fields worth recording, the prompt, the run window and the sheet are all written out in our step by step guide to measuring AI visibility, so they are not repeated here. What follows is the part that comes after the routine: the weekly checks worth running, the conditions that justify a subscription, and the questions to ask before you pay for anything.
Ten minutes a week catches a step change, which is what you are looking for. A monthly run tells you where you are. A weekly run tells you something moved.
Keep a second sheet with the five questions that matter most, usually the ones where you were named but not recommended. Run them on the same weekday, in the same wording, against the same engines.
What a Weekly Sheet Catches
- A competitor published a treatment page you do not have, and the engine now quotes it.
- A directory or review profile added the hours, services or reviews that made the competitor quotable.
- An engine update changed how it handles your city, and the names in your answers shifted.
- Your own site changed and broke a page an engine was reading.
What it cannot do is measure you. Five questions will not show slow drift, and one answer can be an outlier. Treat a weekly result as a reason to run the full set, never as a number to report.
When a Paid Tool Earns Its Price
Twenty questions, four engines, three runs each, one window a month is 240 answers, a solid block of a morning. For one practice that is a fair price for knowing where you stand.
The Conditions That Justify a Subscription
- You are tracking more than a handful of practices, so an agency or a group buying per location.
- You have several locations and need answers per location, not per group.
- Your question set has passed about forty questions.
- You need history: every raw answer kept for months so a change can be dated, not guessed at.
Swipe the table sideways to see every column.
| Question | Manual Spreadsheet | Paid Tool |
|---|---|---|
| Setup | One sheet you build once | Configuration and onboarding per practice |
| Monthly effort | About an hour of your own time | Minutes of checking once it is set up |
| Scope | One practice, one city | Many practices, many cities |
| History | Whatever you keep, in your own files | Depth depends on the plan you buy |
| Raw answers | Yours to export and keep | Only if the tool archives them |
| Comparison | Manual, against the competitors you pick | Built in, for a set you may not choose |
| Effect on visibility | It measures, it does not fix | It measures, it does not fix |
The last row is the one to argue about with a vendor. A tracker tells you the practice is missing from six answers. It does not write the missing page, correct the hours or answer the review that would make the practice quotable. The fixing is separate work.
What To Look For In a Paid Tool
Five Questions To Ask Before You Pay For Anything
- Does it send my exact wording? A tool that paraphrases your prompt is measuring its own question. Ask to see the exact string sent to each engine.
- How many runs per engine per question? One run is a sample of one, and one sample is noise. Three is the smallest number that tells you anything.
- Will it show me the raw answers and let me export them? This decides most comparisons. A vendor who will not show the text behind a number wants you to trust what you cannot verify.
- Can it run the same set for named competitors? Comparison against the practices you lose to is what changes decisions. A tool that only measures you gives you a temperature.
- Does city targeting work the way I think? How the city is injected, whether location is neutralised, and whether a signed in session is used, because a personalised session answers differently.
Three Failure Modes To Watch For
- No archive. A chart of a number with nothing kept behind it. When you want to prove a change, there is nothing to prove it with.
- No engine parity. It claims four engines, but quietly runs one. Ask for the per engine breakdown rather than one combined score.
- Fabricated sentiment scores. A sentiment number with no stated method, no sample and no sentences behind it is decoration. Sentiment means something only if you can read the text.
Before you subscribe, read a real report end to end. drsmiles.ai publishes a sample report with the answers and the scoring visible, which is the level of detail any tool should meet.
How To Keep The Data Clean
Six habits keep a tracker honest. They sound small, and they are the difference between a measurement and a diary.
- Run every question inside the same window each month. If one slips, mark it as a late run rather than pretending it was on schedule.
- Never delete a run that flatters nobody. A missing answer is a gap in the record, and gaps get quietly filled with the run that flattered you.
- Log every change to the question set: the date, the reason, the old wording and the new one. A reworded question starts a new series.
- Keep the scoring rule written down and apply it the same way every month. If it changes, record the date.
- Record failures as failures. A refusal, a timeout, a sign in wall and an empty answer are all rows in the sheet, not blanks.
- Give the tracker an owner: one person, one hour a month. Unowned measurement stops inside two cycles.
That is the whole system: a fixed question set, a fixed window, a sheet and an archive of every answer. The payoff is the month you can point at the answer where a competitor was named instead of you. You can run a first set yourself in about an hour with the snapshot on our home page, which runs ten questions across three engines, then compare it against our benchmark guide for dental practices.