How we measure
One number.
Audited monthly.
Answer Share is the percentage of your golden query set where your practice is named or cited in the generated answer. Here is exactly how it is produced, including the parts designed to stop us from flattering ourselves.
The measurement, in numbers
- Queries, frozen at kickoff
- 40
- Engines swept
- 5
- Runs per query, per engine
- 3
- Observations per month
- 600
- Losses reported alongside wins
- All
- Engine outages absorbed silently
- 0
The protocol
Six hundred observations a month.
The golden set
Forty queries written the way your patients and clients actually type, not the way a keyword tool renders them. Agreed with you in writing at kickoff and then frozen. Changing the questions mid-program is how agencies manufacture improvement.
Five engines
ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. Citation behavior differs enormously between them, and only about eleven percent of domains cited by one are cited by another. A single-engine report describes one vendor's retrieval stack, not your market position.
Three runs per query per engine
Six hundred observations a month. Answer engines are non-deterministic: the same question can return different sources on consecutive runs. Three runs is the minimum that separates a real move from variance.
Named or cited, counted the same way every time
A query counts if your practice is named in the answer text or appears in the cited sources. One definition, applied identically across engines and months, published in your reporting appendix so you can audit it.
Losses and competitors recorded
For every run you did not win, we record who was named instead. Over a quarter that becomes the most useful competitive intelligence in your market, and it is the half of the data most agencies never show you.
Coverage gaps shown, never absorbed
If an engine is unreachable during a sweep, the report shows a gap. It never quietly reduces the denominator, which is the oldest trick for producing a rising line out of a flat one.
Answer Share, twelve months
Share of a 40-query golden set where the practice is named or cited.
- ChatGPT4% → 38%
- Perplexity11% → 62%
- Google AI Overviews7% → 40%
- Gemini3% → 31%
- Claude5% → 35%
| Engine | Month 1 | Month 12 |
|---|---|---|
| ChatGPT | 4% | 38% |
| Perplexity | 11% | 62% |
| Google AI Overviews | 7% | 40% |
| Gemini | 3% | 31% |
| Claude | 5% | 35% |
Illustrative shape only. These are not client results. When Ibid has client results, they will appear here with the practice named and the query set published.
Standards
What we commit to in writing.
We report against booked revenue
Call tracking, form attribution, and booking events are wired at build time. The monthly conversation is about cases booked and matters opened. Sessions and impressions appear in the appendix, where inputs belong.
The remedy is in the contract
If Answer Share has not improved by month six of a Cited or Recommended engagement, we continue working at no fee until it does, or you leave with no termination obligation. It is a clause, not a slogan.
You own everything
The site, the code, the content, the tracking configuration, the query set, and the full measurement history. On exit it transfers to you in a working state. Hostage-taking is common in this industry and it is disqualifying.
No results claimed before there are results
Ibid launched in 2026. This site shows no case studies, no client logos, and no testimonials, because there are none yet. When there are, they will appear with the practice named and the query set published so you can check them.
Questions
On measurement.
Why not just take a screenshot of ChatGPT naming us?
Because the next run may not. Non-determinism means a single favorable output is indistinguishable from luck. If your current agency is sending screenshots, ask them to run the same prompt three times in front of you.
What if our Answer Share is already high?
Then the audit tells you that and you have spent $4,500 to find out you do not need the engagement. That outcome happens, and we would rather deliver it than build a program on a problem you do not have.
Can we see the raw runs?
Yes. Every observation is retained and available to you, including the prompt, the engine, the run index, the date, and the full answer text. The report is a summary of data you can inspect yourself.
How long before the number moves?
Entity and technical corrections often show up within four to eight weeks because they change what the engines can read about you. Corroboration and coverage work moves on a slower curve, generally three to six months. Anyone quoting you a faster number for the second kind is guessing.
The baseline
Start with the number, not the pitch.
The Answer Audit produces your baseline Answer Share and the written reason behind it. $4,500, three weeks, credited in full against any engagement started within sixty days. Ibid is based in San Diego and works with practices nationally.