Finding / 2026-07 to 2026-08 / uk

AI brand shortlists tightened in August

Brand-heavy commercial answers persisted, but the average response made room for slightly fewer tracked brands.

Published
2026-09-02
Revised
2026-09-02
Author
Cleotic Research
Method
comparison-v1.1
2.99

tracked brands per August answer

Denominator: 136 prompts paired across four models and six projects

94.2%

August answers naming a tracked brand

Denominator: 16,126 successful model responses

-0.8%

change in tracked-brand breadth

Denominator: paired July to August prompt comparison

August answers carried fewer brand slots

Each bar indexes July to 100. Tracked-brand presence stayed high, while answers became shorter and the average tracked-brand shortlist narrowed.

United Kingdom / July to August 2026 / Denominator: 136 prompts paired across four continuously observed model surfaces and six projects / Method: comparison-v1.1

View chart data as a table
MeasureJulyAugustChange from July
Tracked brands per answer3.012.99-0.8%
Answers with a tracked brand94.3%94.2%-0.1%
Response length1759 chars1735 chars-1.3%

The available shortlist became a little narrower

The average answer moved from 3.01 tracked brands in July to 2.99 in August, a 0.8% decline. At the same time, the share of answers containing at least one tracked brand slipped by only 0.10 percentage points. Commercial answers remained brand-heavy; they simply carried fewer names on average.

Response length also fell, from 1759 to 1735 characters, while output tokens rose 1.3%. Those measures do not show that answer quality fell. The defensible reading is narrower tracked-brand coverage under this collection design.

Generic visibility is a weak operating target

When an answer has fewer brand slots, broad awareness work is harder to connect to a measurable outcome. The practical unit is the prompt cluster: a use case, comparison, capability, buyer type or decision criterion where a brand should plausibly belong.

Teams should review the prompts where a brand repeatedly misses the shortlist, then compare the brands that do appear and the selection criteria around them. This turns a vague goal such as improve AI visibility into a testable question about eligibility for a specific buying situation.

What marketers should do

  1. 01Build reporting around priority use cases and comparison questions, not one overall mention target.
  2. 02Document the selection criteria repeated across those answers and show evidence against each one.
  3. 03Track who occupies the limited shortlist slots and which selection criteria recur.
  4. 04Retest the same prompt panel after material content or authority work.