Day 0 of 30, 14 Oct 2026
Day 0: our own record, as five engines state it
This is Broadcastwell's own record. We measure other companies' AI records for a living, so we start with ours, in public, under the same rules we sell.
In our State of GEO study, 30 of 85 B2B software companies were never named in the answers for their own category. We show our own result here and fix it in public for 30 days, with a receipt for every step.
Day 0 was measured on 14 October 2026 under the published method, v1.1: ten buyer questions for the category "AI search visibility agencies for B2B software companies", asked on ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode, three scheduled runs, 150 planned answers.
- Named: 0 of 105 valid answers name Broadcastwell (0%, 95% interval 0% to 3.5%).
- Cited as a source: 0 of 105 valid answers cite a page on broadcastwell.com (0%, 95% interval 0% to 3.5%).
- Record accuracy: 0 statements classed wrong or stale. No answer names us, so there is no statement about us to class yet.
45 of the 150 planned answers are excluded from both sides of every count. Each is a Google answer that introduced a list of agencies that did not reach us whole: in 41, Google did not return the list itself; in 2, the answer named none of the agencies it introduced; in 2, advice came where the agencies should have been. We never count an answer we could not read in full as "not named". ChatGPT, Claude and Perplexity returned all 90 of their answers in full; no answer was cut short and no engine returned an error. The 105 valid answers are the base for every rate.
The engines name other agencies instead, Omniscient Digital, WebFX and First Page Sage most often, and cite their pages and third party guides. Every number above links to its receipts.
The plan lists the questions we lose, the sources the engines cite where we are absent, and the changes we will make on our own site and beyond it. Each weekday we publish what we changed. On day 30 we measure again on the same ten questions, the same engines and the same three runs, and publish the before and after, whatever it shows.
No result is promised. We promise the work and the day 30 measurement.
What is next
- Send the first requests to the agency lists the engines cite most
- Publish the first answer page on broadcastwell.com
- Friday: one observation run, labelled one run, not the measurement