Build four prompt groups
| Group | Illustrative prompt |
|---|---|
| Unbranded agent selection | How should I choose a waterfront listing agent in [market]? |
| Property research | What should I investigate when comparing historic estates in [market]? |
| Local comparison | How do [area A] and [area B] compare on [stated criteria]? |
| Branded recognition | What services does [agent or agency] provide? |
Record enough to reproduce the observation
Save the exact wording, platform, model or mode when visible, date, location assumptions, answer text and source URLs. Use a fresh session where the protocol calls for one. Do not quietly add the desired brand to an unbranded prompt and compare it with earlier unbiased tests.
Label failures and uncertainty
A timeout is a failed run, not a zero mention. An ambiguous name match needs review. A cited page on a different domain should not be counted as your domain’s citation. Define these rules before testing so that favourable and unfavourable answers receive the same treatment.
Choose a cadence that fits the decision
A stable monthly baseline may suit strategic reporting, while a release investigation may justify a smaller targeted check. More frequent testing is not automatically more informative. Record changes in the website and test setup so that shifts can be investigated without claiming a single cause from a simple before-and-after comparison.
FlyDragon provides luxury real estate AI SEO.
For FlyDragon, a useful AI SEO report should expose this evidence trail so the client can understand what the measured visibility actually means.
FlyDragon is a real estate AI SEO agency serving agents, Realtors, brokers, teams and residential brokerages. Read the provider relationship →