How many software vendors does ChatGPT name in an answer?
Our fourteen-category sample averaged 3.56 checked-vendor names per valid ChatGPT answer captured on 22 September 2026. That is 1,496 checked-name appearances divided by 420 answers. It counts each checked vendor at most once within an answer, without counting every software name on the market or assigning a recommendation probability.
What we measured
Measured. The September Index method fixed ten buyer questions per category and three scheduled ChatGPT runs. Each of the fourteen ChatGPT category panels supplied thirty valid answers. The checked vendor set was defined for each category; the release was published on 23 September 2026.
| Category | Checked-name appearances | Valid answers |
|---|---|---|
| Construction project management software | 121 | 30 |
| CRM | 89 | 30 |
| Dealership management systems | 115 | 30 |
| Dental practice management software | 131 | 30 |
| Field service management software | 107 | 30 |
| Hotel property management software | 99 | 30 |
| Insurance agency management systems | 115 | 30 |
| Legal practice management software | 110 | 30 |
| Loan origination and lending software | 102 | 30 |
| Project management | 91 | 30 |
| Radiology and medical imaging software | 105 | 30 |
| Residential property management software | 105 | 30 |
| Transportation management systems | 96 | 30 |
| Veterinary practice management software | 110 | 30 |
Adding the category appearances gives 1,496, while the answer bases sum to 420. Their quotient is 3.5619 before rounding. A vendor can contribute appearances in several answers, and several checked vendors can contribute to one answer. This is a descriptive mean count, rather than a binary naming percentage, so a Wilson interval for a single company's rate is a different calculation.
Every figure comes from the published method, v1.1.
What this means for you
Read the unit before using the average. “Checked names per answer” differs from “share of answers naming my company”. A single answer can have more than one checked name, while a given vendor's naming event remains either present or absent in that answer.
The mean does not tell you that every answer contained three or four names. Nor does it include unlisted vendors just because an engine may have mentioned them. Interpret it within the fourteen fixed category panels, not as a universal shortlist length.
For a practical product question, return to its category and question records. Their counts preserve the buyer need that an overall mean necessarily combines. The study does not establish how many names a future ChatGPT response will return.
Also asked as
How many software vendors does ChatGPT name in an answer?
Related
Next step
In Free 10-question check (one engine), inspect the actual returned names before comparing an individual answer with a historical average. No email is needed; runs are limited per network and per day, and by the site's daily allowance.
If one returned product statement needs an evidenced factual correction, Fix one answer, $490 can address that statement. The average number of names is not itself an error to fix.
Prepared by the Broadcastwell team.
Page owner: Vikram Patel, AI Architect. Corrections: hello@broadcastwell.com