How does it differ from other AI visibility measures?
Several related measures are often mixed up. The table separates them. A brand can score high on one and low on another.
| Measure | What it counts |
|---|---|
| Answer inclusion rate | Answers that name the brand in the answer text |
| Citation share | How often the brand's pages are cited or linked as sources, compared with all citations |
| Recommendation rate | How often the brand is included when the engine suggests or shortlists options |
| Mention sentiment | Whether mentions are positive, neutral or negative. See mention sentiment in LLMs |
| Brand co-occurrence | Which other brands appear in the same answers |
How is it calculated?
The basic formula divides the number of answers that name the brand by the total number of answers collected, then multiplies by 100. The example below uses made-up numbers for illustration.
| Item | Value |
|---|---|
| Questions in the set | 50 |
| Runs per question per engine | 3 |
| Answers collected on one engine | 150 |
| Answers naming the brand | 45 |
| Answer inclusion rate on that engine | 30% |
How should the measurement be designed?
- Build a question set that reflects real buyer research, covering broad, comparison and decision questions. See prompt audit.
- Run each question several times on each engine, since answers vary.
- Report results per engine, and a combined figure only as a summary.
- Count name variants, product names and common misspellings, and define whether executive names count.
- Record whether each mention is positive, neutral or negative, and note competitors named in the same answer.
- Keep the question set and method fixed over time, and record changes, so trends are comparable.
Illustrative coding rules (example only)
- Count a mention only if the brand name appears in the answer text
- Count a link or footnote only under citation share
- Code each mention as positive, neutral or negative
- Record other brands named in the same answer
What are the limits?
Inclusion rate is an estimate from a sample. Answers can differ between runs, accounts, locations and dates, and engines change often, so small changes in the rate may be noise. The question set also shapes the result, since a broad set and a narrow set give different numbers. A mention is not an endorsement, and a high rate with negative sentiment is a warning. Tools from different vendors count differently, so compare like with like. No outcome is guaranteed.
What are the common mistakes?
- Mixing up mentions and citations.
- One run per question, which hides variation.
- Missing name variants and misspellings.
- Counting negative mentions as positive presence.
- Ignoring which competitors appear alongside the brand.
- Changing the question set and treating the change as a trend.
Frequently asked questions
What is answer inclusion rate?
It is the share of AI answers in a defined sample that name a brand in the answer text.
Is it the same as citation share?
No. Citation share counts source attributions such as links. Inclusion rate counts mentions inside the answer text. A brand can do well on one and poorly on the other.
What is a good answer inclusion rate?
There is no standard. It depends on the category, the question set and the competition. Compare with your own past results and with close competitors measured the same way.
How many runs are needed?
Enough to see the variation, usually several per question and engine. More runs give steadier estimates.
Can a brand raise its inclusion rate?
Possibly, by making accurate information available and earning credible independent coverage, but engines choose on their own and results should be re-tested. See visibility gap analysis.
By the 5W Editorial Team. Reviewed by a human editor before publication. Updated October 28, 2026.