ChatGPT is naming fewer brands when it answers B2B technology questions than it did in March, according to a new BrightEdge study. The narrowing hits hardest at the top of the funnel, where buyers are still forming a shortlist rather than weighing named vendors against each other. That makes each surviving mention more valuable, and considerably harder to earn.
BrightEdge, the enterprise SEO platform, built the finding using its AI Catalyst tracking tool: a fixed panel of non-branded B2B Tech prompts run on ChatGPT and held identical across two dates, March 15 and September 13, 2026. Across every funnel stage combined, the average count of distinct brands per answer dropped from 5.36 to 3.87, a 27.9 percent decline. The study measures mentions, not citation counts or market share, and it covers ChatGPT alone. BrightEdge says whether Google’s AI Overviews or AI Mode show a similar pattern has not been checked.
The decline was not even across the funnel. Informational prompts, the broad opening questions a buyer asks before naming any vendor, fell from 5.15 to 3.50 brands per answer, a 32.0 percent drop and the sharpest of the stages tracked. Consideration prompts, where a buyer is weighing named options against each other, fell from 8.21 to 6.39, down 22.1 percent. Transactional prompts moved the opposite way, climbing from 2.02 to 3.34. BrightEdge excluded that figure from its conclusions because a third, unpublished data point shows the direction reversing.
Among informational answers, presence dropped hardest for Microsoft, down 9.5 percentage points, followed by Google, down 6.7 points, and Amazon, down 6.4 points. Several developer and data-infrastructure names, including Apache, dbt, Snowflake, Elastic, and HashiCorp, slipped alongside them. IBM gained more ground than any other brand, up 9.9 points, and government and standards sites such as NIST.gov and FTC.gov climbed too.
A similar split showed up in consideration-stage answers. GitLab, Apple, Trend Micro, GitHub, Amazon, and Google each lost share there, while every brand that gained was a narrow, single-purpose tool rather than a horizontal platform. Zoho’s presence rose from 8.5 percent to 12.8 percent, the largest gain BrightEdge recorded at that stage. Wave Apps, SAP, and a run of security and infrastructure names posted smaller increases.
BrightEdge’s own numbers complicate a simple read on why the list is shrinking. At the informational stage, the number of sources cited per answer climbed even as named brands fell, moving from 2.30 to 3.12, so fewer brands were named but each was backed by more sourcing. At the consideration stage, citation counts declined alongside brand mentions, moving from 5.70 to 3.25. That split matters because it undercuts an easy publisher-squeeze story. TechTarget gained informational-stage presence over the same six months, and BrightEdge states directly that the pattern does not support a story about publishers losing ground to owned content.
This is one vendor’s fixed-panel read of one AI engine in one vertical, comparing two snapshots six months apart. It is not a verdict on AI search broadly or on how ChatGPT users behave outside B2B tech. B2B marketers should still treat it as a prompt to check ChatGPT visibility stage by stage rather than by a single blended score, since an absence from informational answers and an absence from consideration answers call for different fixes, and the awareness-stage shortlist just got shorter for everyone competing on it.
Per BrightEdge’s AI Catalyst research, published September 21, 2026.
