BrightEdge tracked the same set of B2B tech questions in ChatGPT in March and September 2026. The average answer named 5.36 brands in March and 3.87 in September, 27.9% fewer. The steepest fall came in early questions, where buyers first learn about a category.
Table of contents
Key takeaways
- On a fixed panel of non-branded B2B tech prompts, ChatGPT named 27.9% fewer brands per answer in September than in March 2026.
- Informational questions fell most, from 5.15 to 3.50 brands. Comparison questions fell from 8.21 to 6.39.
- Big platforms lost ground in comparison answers. Single-purpose tools such as Zoho and Wave gained.
- BrightEdge publishes no prompt counts or panel size, so the reliability of these averages cannot be checked.
What this is about
BrightEdge, the SEO platform, published this in its weekly AI search report on September 21, 2026. It compares two dates six months apart on the same prompts. For B2B brands, the list of names in a ChatGPT answer got shorter, so each place on it is harder to win.
What the source says
The measure is simple: how many different brands ChatGPT names in one answer, averaged across a group of prompts. BrightEdge stresses that this is not citation volume and not market share.
The prompts were grouped by buyer stage. Informational prompts ask what a technology is. Consideration prompts compare options, such as accounting software for a small business. Transactional prompts come when someone is ready to buy.
BrightEdge held one stage back from its conclusions. Transactional answers rose from 2.02 to 3.34 brands, but a third check showed that move reversing. BrightEdge printed both numbers and said it would wait for more data. That is good practice, and rarer than it should be.
What the data shows
| Buyer stage | March 2026 | September 2026 | Change |
|---|---|---|---|
| All stages | 5.36 | 3.87 | -27.9% |
| Informational | 5.15 | 3.50 | -32.0% |
| Consideration | 8.21 | 6.39 | -22.1% |
| Transactional | 2.02 | 3.34 | not treated as a trend |
About 15% of prompts moved to a different stage between the two dates. The main figures use only prompts that kept the same stage. Across all prompts the overall fall was 25.1%, close to the 27.9% headline.
In informational answers, Microsoft, Google and Amazon lost the most, by 9.5, 6.7 and 6.4 points. IBM gained 9.9 points. Government and institutional sources such as NIST.gov and FTC.gov also rose, along with the publisher TechTarget.
In comparison answers, Google lost 9.3 points and Amazon 6.1. Zoho rose from 8.5% to 12.8% of answers. Wave Apps went from 2.8% to 5.7%, and SAP from 3.5% to 5.7%. BrightEdge notes that every brand gaining here is a single-purpose tool.
Citations did not move with mentions. In informational answers, fewer brands were named but citations per answer rose, from 2.30 to 3.12. In comparison answers both fell, citations from 5.70 to 3.25.
On the seven method points, BrightEdge states the engine, region, dates and prompt type. It gives no panel size and no prompt counts, on purpose, to keep client data private. It says nothing about repeated runs. With no count, a four-point gain for Zoho could rest on a handful of answers. The two-date design also carries the risk BrightEdge itself flagged for transactional prompts: a third date might tell a different story.
What it does not mean
It does not mean publishers are being pushed out. BrightEdge checked this. TechTarget gained, and other large publishers held steady.
It does not mean fewer sources. In informational answers, citations per answer went up while brand names went down.
It does not apply to Google. This is ChatGPT only. BrightEdge says AI Overviews and AI Mode are untested, and that Google is still the bigger channel.
It is not market share. A brand named in 12.8% of answers is not 12.8% of anything else. For a related pattern, see what we found when AI does not know a brand.
What to do
1. Split your tracked prompts by buyer stage. One visibility score hides the stage you are missing from. 2. Find the comparison questions you can win. The brands gaining ground own one clear use case. Name yours and answer that exact question on your site. 3. Keep early-stage content, but aim it narrower. Informational answers now name fewer brands. Pick one to three questions where you should appear but do not. 4. Check again in a quarter. Two dates show a direction, not a trend. Our guide to how answer engines cite explains why answers shift.
FAQ
Why would ChatGPT name fewer brands?
BrightEdge does not say, and the data cannot show a cause. It only shows that answers on the same prompts got shorter lists over six months.
Does this hold outside B2B tech?
Unknown. This edition covers B2B tech in the US on ChatGPT only. Other categories and engines were not part of it.
Why was the buying stage left out?
Its numbers rose between March and September, but a third check showed them falling back. BrightEdge chose to wait for the full weekly series rather than report a move that might not last.
