What does our ChatGPT visibility work include?
Broadcastwell measures and fixes whether AI search engines name B2B software companies.
Run the free checkOur ChatGPT visibility work is the five-engine work every offer runs, with ChatGPT read on its own line: we record what ChatGPT answers to your buyer questions, find the page or source behind a missing or wrong answer, repair it with facts you approve, and ask again. No offer measures ChatGPT alone, and nothing we do can make ChatGPT name you.
Updated 11 November 2026.
Every capture covers ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode with identical wording, so a ChatGPT result always sits beside four others. The figures below come from the September 2026 Absence Index: 140 fixed buyer questions across fourteen software categories, three scheduled runs each.
What ChatGPT did in the September Index
| Measure, September 2026 | ChatGPT | All five engines |
|---|---|---|
| Valid target answers | 420 of 420 attempts | 2,095 of 2,100 attempts |
| Cited entries | 1,993 | 18,893 |
| Answers with no cited entry | 0 of 420, 0.0% (0.0% to 0.9%) | 39 of 2,095, 1.9% (1.4% to 2.5%) |
| Vendor sites in cited entries | 1,708 of 1,993, 85.7% (84.1% to 87.2%) | 5,766 of 18,893, 30.5% (29.9% to 31.2%) |
| Review platforms in cited entries | 5 of 1,993, 0.3% (0.1% to 0.6%) | 1,706 of 18,893, 9.0% (8.6% to 9.4%) |
| Checked-vendor names | 1,496 | 8,085 |
| Vendor-question cells named in one or two of three runs | 159 of 2,860, 5.6% (4.8% to 6.5%) | 1,259 of 14,192, 8.9% (8.4% to 9.4%) |
Measured. ChatGPT leaned on vendor sites more than any other engine: 85.7% of its 1,993 cited entries went to the websites of software vendors, against 30.5% for the five engines together. Review platforms supplied only 5 entries. Government and standards bodies, a rare class elsewhere, supplied 43 (2.2%), and 11.2% fell in our "other" class. Every one of the 420 valid ChatGPT answers carried at least one cited entry, and ChatGPT had the lowest share of vendor-question cells that changed between its three runs.
Who ChatGPT named most in each category
| Category | Valid ChatGPT answers | ChatGPT checked-vendor names | Named most by ChatGPT | Top three share of ChatGPT names |
|---|---|---|---|---|
| Construction software | 30 | 121 | Procore, Buildertrend, Contractor Foreman | 45.5% (36.9% to 54.3%) |
| CRM software | 30 | 89 | HubSpot CRM, Salesforce Sales Cloud, Pipedrive | 69.7% (59.5% to 78.2%) |
| Dealership software | 30 | 115 | DealerCenter, CDK Global, Dealertrack | 40.0% (31.5% to 49.1%) |
| Dental software | 30 | 131 | Dentrix, Open Dental, Dentrix Ascend | 49.6% (41.2% to 58.1%) |
| Field service software | 30 | 107 | Housecall Pro, ServiceTitan, Jobber | 54.2% (44.8% to 63.3%) |
| Hotel software | 30 | 99 | Cloudbeds, Mews, Stayntouch | 59.6% (49.7% to 68.7%) |
| Insurance agency software | 30 | 115 | Applied Systems, EZLynx, Vertafore | 59.1% (50.0% to 67.7%) |
| Legal practice software | 30 | 110 | Clio, 8am MyCase, PracticePanther | 52.7% (43.5% to 61.8%) |
| Loan origination software | 30 | 102 | MeridianLink, nCino, ICE Mortgage Technology | 43.1% (33.9% to 52.8%) |
| Project management software | 30 | 91 | Asana, Jira, ClickUp | 53.8% (43.7% to 63.7%) |
| Property management software | 30 | 105 | Buildium, AppFolio, DoorLoop | 53.3% (43.8% to 62.6%) |
| Radiology software | 30 | 105 | Intelerad, RamSoft, Sectra | 37.1% (28.5% to 46.7%) |
| Transportation software | 30 | 96 | Descartes, Descartes Aljex, Alvys | 32.3% (23.8% to 42.2%) |
| Veterinary software | 30 | 110 | ezyVet, Shepherd, Digitail | 51.8% (42.6% to 60.9%) |
Our industry benchmarks pool the five engines for each category; the table above keeps ChatGPT apart. Where ChatGPT's most-named vendor differs from the pooled one, the difference is an observation on those dates, not a verdict on either product.
What each offer gives you for ChatGPT
| Offer | ChatGPT in the deliverable |
|---|---|
| Live question, $19 once | Three of the 15 planned answers are ChatGPT runs of your question, each with its receipt, within 24 hours of payment |
| Answer Repair, $490 once | The five-answer record at Ready holds one ChatGPT answer with every URL it cited; the repair goes live or ships as a handoff within 48 elapsed hours of Ready |
| Agency Deadline Audit, $490 per client report | 30 of the 150 observations are ChatGPT answers to the ten agreed questions |
| Fix Sprint, $3,900 ($3,410 after the $490 credit) | ChatGPT has its own line in the day 0 and day 30 evidence for the agreed test |
| Care, $490 a month | A monthly ChatGPT line beside the other engines, and work on the outside pages they cite for your category |
| Program, $5,500 a month | ChatGPT figures in the separate monthly measurement report |
Delivery clocks and refund terms are quoted exactly in our operating framework; the service model shows the workstreams behind each row.
OpenAI changes we track for ChatGPT
OpenAI's ChatGPT release notes, read 9 November 2026, dated six model entries between 6 August and 26 October 2026. The entry dated 22 September 2026, our first capture day, was scoped to ChatGPT Work and Codex, which OpenAI describes as separate from the models in Chat. The 6 August entry concerned the chat experience before our capture, and the next Chat tab change is dated 7 October 2026, after it. Our model release tracker lists each one with its scope, and we never credit a change in a ChatGPT count to a release.
What this means for a ChatGPT repair
Reasoned. Because 85.7% of ChatGPT's cited entries were vendor sites, start with your own site when ChatGPT leaves you out or gets a fact wrong: the product, pricing and comparison facts a buyer question needs, stated plainly on pages a reader can open. A three-run ChatGPT miss was also the steadiest pattern in this record, so a repeated absence is worth investigating rather than waiting out. Neither point predicts the next ChatGPT answer.
Sources
- Broadcastwell, The Absence Index, release 2026-09, release record, DOI 10.5281/zenodo.22907695; target answers captured 22 and 23 September 2026, published 23 September 2026.
- Broadcastwell, ChatGPT rows of the answer citation bases, post-audit source classes, repeat disagreement, vendor concentration and excluded answers tables, report DOI 10.5281/zenodo.23804758. The names total sums the fourteen ChatGPT rows of the concentration table.
- OpenAI Help Center, ChatGPT release notes, entries dated 6 August to 26 October 2026, read 9 November 2026.
- Broadcastwell, Pricing and Methodology, read 9 November 2026.
Related: our ChatGPT answers, why each engine is read separately, and the study of who appears in the sources of B2B software answers.