How do we compare citation metrics when changing AI tools?
Match the event being counted and its valid base before comparing the rates. In answers captured on 22 September 2026, Housecall Pro was named in 108 of 150 answers, while its own domain was cited in 43 of 150. Those results give you a naming rate of 72.0 percent and a domain-citation rate of 28.7 percent from the same collection; changing which field your report shows does not establish a change in the answers.
What we measured
Measured. The Absence Index, release 2026-09, published 23 September 2026 under DOI 10.5281/zenodo.22907695, records naming and own-domain citation separately. The field service category used ten fixed buyer questions, asked three times on ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode. Housecall Pro's record retains 150 valid answers from 22 September 2026.
| Housecall Pro measure | Count and base | Rate | Wilson 95 percent interval |
|---|---|---|---|
| Vendor named in the answer | 108 of 150 | 72.0 percent | 64.3 to 78.6 percent |
| Vendor's own domain cited | 43 of 150 | 28.7 percent | 22.0 to 36.4 percent |
The citation figure counts eligible answers citing Housecall Pro's domain. It does not give the share of all cited URLs or all citation events. Those would require their own denominators. These observations describe one dated collection, not a test of either tool you are considering.
What this means for you
Reasoned. Build a field map before carrying an old score into a new report. For this example, the useful questions are concrete:
| Field | What to match |
|---|---|
| Event | Does the column count a vendor name, a cited domain or something else? |
| Base | Does 150 mean valid answers, attempted answers or citation events? |
| Unit | Does the export store 72.0 percent as 72.0 or as 0.72? |
| Collection | Are the questions, engines, dates, repeat counts and exclusions comparable? |
| Prominence | What position or weighting does the tool actually measure, and where is its definition? |
Dentrix shows why the base deserves its own check. Its dental record retained 149 valid answers after one exclusion on 22 September 2026. It was named in 120 of them: 80.5 percent, with a Wilson 95 percent interval of 73.4 to 86.1 percent. Dividing 120 by 150 instead produces 80.0 percent. That denominator decision changes the calculation without supplying another observed answer.
The September release supplies no prominence score. We cannot derive one from its naming and citation columns or use them to explain another platform's formula. Keep that field unmatched until its definition and supporting records are available. A label such as "citation share" needs the same check; the name alone does not tell you its denominator.
- If you report for an agency: keep the old definition beside the new one. Mark a changed measure as a new series rather than presenting it as a client's gain or loss.
- If you sell software: ask which buyer questions produced the comparison. Adding an engine or changing the questions changes what the report covers.
- If you buy a tool: request a sample export with its field definitions. A visible number without its event and base cannot resolve this comparison.
Also asked as
How do we compare citation metrics when changing AI tools?
Related
- Housecall Pro's field service record
- Dentrix's dental practice record
- Why naming and citation differ
- Scoring and matching definitions
- What a valid measurement requires
- Read the cross-category score example
Next step
Use the Free 10-question check (one engine) to inspect a bounded result. No email is needed; runs are limited per network and per day, and by the site's daily allowance. Keep that check's engine and questions beside its result when comparing it with a broader report.
If the evidence identifies a particular page or fact to correct, Fix one answer, $490. A completed source-backed change has a defined deliverable; a higher prominence score is not its acceptance condition.
Every figure comes from the published method, v1.1.
Claiming or correcting a vendor record is free and never depends on a purchase.
Prepared by the Broadcastwell team.
Page owner: Vikram Patel, AI Architect. Corrections: hello@broadcastwell.com