All insights

AI answer audits

When a competitor appears in the answer about your own practice

Search your practice name in an assistant. If another office shows up in the reply, the problem is entity confusion, not competition.

A grid of tracked questions scored across answer engines

The owner’s takeaway

A branded question should return the practice, its clinicians, and its location. When a competitor appears instead, the cause is usually a shared address, a similar name, or a directory record that merged two businesses.

Type a practice name into an assistant and read the reply. Most owners expect their own site to be summarised. A surprising number get a paragraph that mentions a nearby competitor, an old location, or a dentist who left in 2022.

This is not a ranking loss. It is an identity problem. The model is not sure which entity the name refers to, so it widens the answer to include businesses that share a signal.

Identity before marketing

  1. Test

    Ask the brand question

    Three runs, two engines, saved verbatim.

  2. Count

    Foreign names

    Any competitor mentioned is a disambiguation failure.

  3. Repair

    One canonical record

    Name, address, phone, current clinicians.

  4. Verify

    Repeat after 30 days

    Same question, same engines.

A competitor in a branded answer is a data problem with a marketing symptom.

Four causes worth checking first

  • Shared address: two practices in one building, or a suite change the directories never updated.
  • Similar names: “[City] Dental” and “[City] Dental Care” are one token apart to a model.
  • Merged profiles: an acquisition left two listings, two phone numbers, and two sets of hours.
  • Departed clinicians: a bio page still lists an associate who now works across town.

Each cause produces the same symptom: the answer hedges, then includes a second practice to be safe. The fix is disambiguation, not more marketing.

Disambiguation is a publishing job

A practice needs one canonical identity: legal name, trading name, address, phone, and the clinicians who currently work there. Every profile that carries a different version becomes a competing claim. The work is unglamorous, and it is the highest-return repair for branded queries.

On the site, an organization block with the same name, address and phone as the profile, a clinician page per provider with credentials, and a location page that states which services happen there. When these agree, the model stops hedging.

How to measure whether it worked

Run the branded question three times on two engines, before and after. Count three things: whether the practice is named, whether any other practice is named, and whether the cited sources are the practice's own pages. A repair shows up as fewer foreign names and more first-party citations, often before the prose improves.

Your next move

Check what your own name returns

The free snapshot includes a branded question alongside five treatment questions, so identity problems surface early.

Free AI snapshotHow the method handles accuracy