Cross-model agreement is the movement worth investigating first
15 tracked brands moved in the same direction on every continuously observed model surface.
- Published
- 2026-09-02
- Revised
- 2026-09-02
- Author
- Cleotic Research
- Method
- comparison-v1.1
brands up on all four model surfaces
Denominator: 100 tracked brands across six projects
brands down on all four model surfaces
Denominator: 100 tracked brands across six projects
model agreement required
Denominator: the continuously observed comparison panel
15 brands moved in one direction on all four models
5 tracked brands gained visibility on every model and 10 declined on every model. The mean bar sorts the cases. Agreement across the four separate measures is the reason to review them.
United Kingdom / July to August 2026 / Denominator: Brand visibility within each project and model; four continuously observed model surfaces / Method: comparison-v1.1
View chart data as a table
| Brand | Sector | Mean change | Claude Sonnet 5 | Seed 2.0 Lite | Gemini 3.5 Flash | Grok 4.3 |
|---|---|---|---|---|---|---|
| By Miles | Insurance | +1.80 pp | +0.00 pp | +1.87 pp | +1.19 pp | +4.12 pp |
| ManyPets | Insurance | +1.15 pp | +1.32 pp | +1.05 pp | +0.74 pp | +1.49 pp |
| Marshmallow | Insurance | +1.01 pp | +1.79 pp | +0.92 pp | +1.14 pp | +0.18 pp |
| EDF Energy | Energy and home | +0.99 pp | +0.22 pp | +1.86 pp | +1.16 pp | +0.72 pp |
| Dext | SME accounting | +0.64 pp | +0.40 pp | +0.34 pp | +1.63 pp | +0.18 pp |
| Sophos | Cybersecurity | -0.44 pp | -0.51 pp | -0.45 pp | -0.08 pp | -0.72 pp |
| GoCardless | Fintech | -0.61 pp | -0.36 pp | -0.33 pp | -0.78 pp | -0.96 pp |
| CrowdStrike | Cybersecurity | -0.79 pp | -1.12 pp | -0.99 pp | -0.84 pp | -0.20 pp |
| Hiscox | Insurance | -0.83 pp | -1.21 pp | -0.50 pp | -0.80 pp | -0.83 pp |
| PayPal | Fintech | -0.86 pp | -0.25 pp | -1.00 pp | -1.40 pp | -0.80 pp |
| Palo Alto Networks | Cybersecurity | -0.97 pp | -1.07 pp | -0.55 pp | -0.74 pp | -1.53 pp |
| Stripe | Fintech | -0.99 pp | -0.15 pp | -1.14 pp | -1.37 pp | -1.29 pp |
| Hive | Energy and home | -1.27 pp | -2.01 pp | -0.15 pp | -1.05 pp | -1.86 pp |
| Crunch | SME accounting | -1.51 pp | -0.76 pp | -1.15 pp | -2.01 pp | -2.11 pp |
| Luminance | Legal tech | -2.15 pp | -2.07 pp | -2.13 pp | -1.08 pp | -3.31 pp |
Direction across models is a stronger triage signal than one large average
By Miles, ManyPets, Marshmallow, EDF Energy, Dext gained visibility on all four model surfaces. Sophos, GoCardless, CrowdStrike, Hiscox, PayPal, Palo Alto Networks, Stripe, Hive, Crunch, Luminance declined on all four. These cases stand out because their direction did not depend on one model.
Agreement does not prove a campaign worked or failed. It helps rank the review queue. A broad movement deserves prompt-level inspection before an equally large aggregate shift caused by one model surface.
The next question is where the movement came from
For each brand, identify the prompts that changed, whether the brand left the answer entirely or moved out of a shortlist, and which competitors took the available places.
Teams should also look for concentration by intent. A movement confined to alternatives prompts calls for a different response from one spread across capability discovery and single-choice decisions.
What marketers should do
- 01Prioritise all-model directional agreement in the investigation queue.
- 02Trace each movement to prompts, competitors and answer changes.
- 03Separate disappearance from a lower position inside the same answer.
- 04Check whether the movement clusters around a specific buyer intent.