| Category | Function | Share | Rank | Negative rate | Labels | Quadrant | Since September 2026 |
|---|---|---|---|---|---|---|---|
| Zero trust network access | Network and edge | 38% | 2 of 66 | 4% | 56 | endorsed leader | ▼−9Since September 2026: 42% → 33%, −9 points. Inside the 11-point floor: within noise. Read over the models both editions asked. |
| SASE platforms | Network and edge | 33% | 2 of 45 | 13% | 62 | endorsed leader | ▲+9Since September 2026: 23% → 33%, +9 points. Inside the 11-point floor: within noise. Read over the models both editions asked. |
| DNS filtering | Network and edge | 4% | 4 of 94 | 0% | 4 | under 10 labels · led by DNSFilter at 59% | |
| Network access control | Network and edge | 2% | 9 of 111 | 0% | 2 | under 10 labels · led by Portnox Cloud at 35% | |
| AI security platforms | Security operations | 0% | 202 of 233 | 50% | 2 | under 10 labels · led by SentinelOne Singularity at 11% | |
| Proxy and web data providers | Data platform | 0% | 39 of 92 | 0% | 1 | under 10 labels · led by Decodo at 41% |
| Category | September 2026 | Now | Change | Reading | Rank |
|---|---|---|---|---|---|
| SASE | 23% | 33% | ▲+9Since September 2026: 23% → 33%, +9 points. Inside the 11-point floor: within noise. Read over the models both editions asked. | Within noise | Rank 3 → 2 of 45 |
| ZTNA | 42% | 33% | ▼−9Since September 2026: 42% → 33%, −9 points. Inside the 11-point floor: within noise. Read over the models both editions asked. | Within noise | Rank 1 → 2 of 66 |
| DNS filtering | 0% | 4% | newNew since September 2026: not ranked then, 4% now. | New this edition | Rank 4 of 94, unchanged |
| NAC | 0% | 2% | newNew since September 2026: not ranked then, 2% now. | New this edition | Rank 9 of 111, unchanged |
| AI security | 0% | 0% | newNew since September 2026: not ranked then, 0% now. | New this edition | Rank 202 of 233, unchanged |
| Proxy and web data | 0% | 0% | newNew since September 2026: not ranked then, 0% now. | New this edition | Rank 39 of 92, unchanged |
Shares here are read over the models both editions asked, so they can differ by a point or two from the standing above, which counts every model in this edition.
The floor is 11 points of share, measured: how far the models move a leader on their own when the same questions are asked twice with nothing changed. A larger change is movement; a smaller one is noise, and both are shown. Movement is read over the twelve models both editions asked; GPT-6 Luna, Muse Glimmer 30B joined this edition and are in the standing but not yet in the comparison. How the floor is measured · The editions
| Model | First choice | Alternative | Mention | Negative | Labels |
|---|---|---|---|---|---|
| Claude Haiku 4.5 | 3 | 4 | 3 | 0 | 10 |
| GPT-5.4 mini | 3 | 2 | 1 | 0 | 6 |
| Gemini 3.5 Flash | 2 | 9 | 1 | 0 | 12 |
| Perplexity Sonar | 2 | 3 | 2 | 1 | 8 |
| Grok 4.20 | 5 | 4 | 1 | 2 | 12 |
| Mistral Small | 1 | 5 | 3 | 0 | 9 |
| DeepSeek V4 Flash | 4 | 5 | 0 | 2 | 11 |
| Llama 4 Maverick | 2 | 1 | 4 | 0 | 7 |
| Qwen 3.7 Flash | 3 | 4 | 0 | 1 | 8 |
| Kimi K2 | 4 | 4 | 2 | 0 | 10 |
| GLM 4.7 FlashX | 2 | 5 | 1 | 1 | 9 |
| MiniMax M2.5 | 5 | 1 | 1 | 1 | 8 |
| GPT-6 Luna | 3 | 1 | 2 | 1 | 7 |
| Muse Glimmer 30B | 4 | 4 | 0 | 2 | 10 |
Verbatim evidence the judge attached to positive labels.
“Pure ZTNA (e.g., Twingate, Tailscale, Cloudflare Zero Trust): Great if you just want to kill the VPN and grant app-level access. Low overhead, easy to implement.” Qwen 3.7 Flash · ZTNA · scale prompt · first choice
“My recommendation: Start with Cloudflare One's free tier... It's the lowest-risk, most transparent SASE option on the market right now” DeepSeek V4 Flash · SASE · budget prompt · first choice
“the strongest practical choice is usually Cloudflare One, because it has the clearest low-cost entry point and even a free tier” Perplexity Sonar · SASE · budget prompt · first choice
“Cloudflare Zero Trust (specifically Cloudflare Access for ZTNA) stands out as the best option for a company with a limited budget.” Grok 4.20 · ZTNA · budget prompt · first choice
Verbatim evidence attached to negative labels. A warning on a product with few labels is a warning; on a product with many, it is one voice among them.
“lacks mature security features ... Avoid if: You need comprehensive security features alongside connectivity” GLM 4.7 FlashX · SASE · negative prompt · hard negative
“Cloudflare Zero Trust are cited as genuinely useful free products, though with limited coverage, no SLA and manual configuration” Muse Glimmer 30B · AI security · budget prompt · soft negative
“recommends testing it rather than assuming it matches the traditional routing, local-services, and survivability depth” Perplexity Sonar · SASE · negative prompt · soft negative
“Complex onboarding, steep learning curve, limited features, opaque pricing (26% negative labels in AI review analysis)” Grok 4.20 · SASE · negative prompt · soft negative
Citations exist only for the models that return a source list, five of the fourteen in this edition, so these counts come from 372 of the 397 answers that named Cloudflare One and are not a share of its labels.
772 of the 862 domain citations in answers naming Cloudflare One came from somebody else's page.
Pages are listed as the models cited them.
Search figures are US estimates from DataForSEO, read September 28, 2026; AI search demand is its modeled, directional estimate, not a count of queries to any assistant. The answers are this edition's. Two measurements side by side: neither is read as the cause of the other.
An email the morning each edition publishes: where this product moved, where it held, and by how much against the noise floor. One address, confirmed by a click; a stop link in every email.
Already following? Everything you follow, with a stop for each.
What Cloudflare One's own pages state, read September 28, 2026: cloudflare.com/enterprise, blog.cloudflare.com/cloudflares-2012, developers.cloudflare.com/cloudflare-one, blog.cloudflare.com/cloudflares-global-reach, blog.cloudflare.com/cloudflare-data-platform. A claimed page can correct any of them.
Claiming is free and changes nothing in the data. A claimed page shows a verified contact who is told when each edition publishes and when Cloudflare One's standing changes by more than the noise floor; the right to propose corrections to the vendor table, meaning names the judge wrote that should or should not read as Cloudflare One, applied by version and listed in the change log; and a one-line description supplied by the vendor and marked as such.
A new claim receives the current edition's vendor brief for Cloudflare One by email, built from the raw record of the edition. It shows:
Cloudflare One is among the products AI models recommend first in Zero trust network access this edition. Each badge says so in the buyer's words, names the edition, and links to the standing. The next edition issues a new badge; this one stays true as a record of October 2026.
Cloudflare One is among the products AI models recommend first in SASE platforms this edition. Each badge says so in the buyer's words, names the edition, and links to the standing. The next edition issues a new badge; this one stays true as a record of October 2026.
Badges read on light and dark pages. Alt text carries the claim, the category, the edition and the index, so it stays a citation where the image does not load.