Zero of fourteen models named Cloudflare Page Shield first on the direct prompt; one named Source Defense. Cloudflare Page Shield was named by twelve of the fourteen models and Source Defense by eleven and Cloudflare Page Shield carries 29 labels and Source Defense 23, so the shares are not directly comparable.
Named in one category this edition.
Named in one category this edition.
Share is the count of first choices across the direct, paraphrase, budget and scale prompts over all fourteen models, for a mid-market B2B company; rank is within the category; every quote names the model and the prompt it came from. Both figures come from the client-side protection page.
Across every category in the November 2026 Edition, Cloudflare Page Shield and Source Defense were named in the same answer thirty-two times, of the 80 answers naming Cloudflare Page Shield and the 59 naming Source Defense. In those answers Source Defense took the first choice two times and Cloudflare Page Shield seven.
| Model | Direct | Paraphrase | Comparative | Budget-constrained | Scale-constrained | Negative |
|---|---|---|---|---|---|---|
| Claude Haiku 4.5 | ||||||
| GPT-5.4 mini | ||||||
| Gemini 3.5 Flash | ||||||
| Perplexity Sonar | ||||||
| Grok 4.20 | ||||||
| Mistral Small | ||||||
| DeepSeek V4 Flash | ||||||
| Llama 4 Maverick | ||||||
| Qwen 3.7 Flash | ||||||
| Kimi K2 | ||||||
| GLM 4.7 FlashX | ||||||
| MiniMax M2.5 | ||||||
| GPT-6 Luna | ||||||
| Muse Glimmer 30B |
Bold names in an answer are the products the judge labeled a first choice; a model naming several gives each of them that label. The full answer text for every row is in the record.
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Six of eight in this category shown.
“Convenient if you already run Cloudflare, but a major migration if you don't. Free tier exists but isn't compliance-sufficient.” DeepSeek V4 Flash · paraphrase prompt · soft negative
“Cloudflare Page Shield samples down to 1% of sessions, leaving 99% of user sessions without monitoring” Claude Haiku 4.5 · comparative prompt · soft negative
“it ranks less strongly for mid-market client-side protection than the leading options” Perplexity Sonar · direct prompt · soft negative
“For most mid-sized B2B companies already on Cloudflare, Cloudflare Client-side Security gives you script inventory, change alerting and PCI 6.4.3/11.6.1 evidence with near-zero ops overhead.” Muse Glimmer 30B · paraphrase prompt · first choice
“Cloudflare Client-Side Security (Page Shield) and Feroot Security are the best choices due to their free tiers and strong reputations.” Mistral Small · budget prompt · first choice
“If you are on Cloudflare: Enable Cloudflare Page Shield. It is the path of least resistance and covers PCI 4.0 requirements natively.” Gemini 3.5 Flash · paraphrase prompt · first choice
Every negative label with a quote, up to three, then the highest-weighted positives, up to three. Six of eight in this category shown.
“Tools that rely solely on client-side JavaScript agents or sandboxing (such as those from vendors like JScrambler, SourceDefense, Reflectiz, and Human Security) are often criticized.” Mistral Small · negative prompt · hard negative
“"such as JavaScript sandboxing agents (e.g., from vendors like JScrambler, SourceDefense/Reflectiz, Human Security)" ... 7 of 14 models · 3 hard negative” Muse Glimmer 30B · negative prompt · hard negative
“Vendors like SourceDefense, JScrambler, Reflectiz (in some implementations), Human Security, and otto-js have been criticized” Kimi K2 · negative prompt · hard negative
“Source Defense is the strongest default choice” Perplexity Sonar · direct prompt · first choice
“Best for organizations that rely heavily on tag managers and need to control the blast radius of third‑party scripts” MiniMax M2.5 · comparative prompt · alternative
“Choose Source Defense when enforcing fine-grained permissions for third-party scripts matters most.” Perplexity Sonar · comparative prompt · alternative
Comparisons are drawn for the top eight products in each category, each against each. The output is the models' output; nothing here is a recommendation by the index.