AI comparison
Don’t Take Our Word for It. Ask AI.
The plan is simple: give leading AI models the same neutral prompt, ask them to compare MailThreatZero with the major email-security platforms an MSP actually evaluates, and publish what comes back without editing it.
No results are published yet
Nothing on this page is an AI evaluation of MailThreatZero, because none has been published. What is here is the exact prompt and the method. When results are published they will appear here in full, with the model and the date, and this notice will be gone.
We would rather show you an empty page than a page of numbers we made up.
About this comparison
Any AI comparison shown here is commissioned and published by MailThreatZero. Each listed model receives the same comparison prompt. Raw responses are retained and displayed without editing except for formatting and removal of secrets or API metadata. This is not independent laboratory testing, certification, or an endorsement by the AI providers.
AI models can make mistakes. Verify important security, pricing, and product claims directly with the referenced vendor.
Same question. Different models.
AI models do not always agree. That is exactly why we show more than one. Compare how ChatGPT, Claude, Gemini, Grok and Perplexity independently evaluate the same products using the same published criteria.
We asked leading AI models to compare MailThreatZero with major email-security platforms using the exact same prompt. No special MailThreatZero prompt. No hidden scoring formula. No edited conclusions. See the responses, inspect the methodology, copy the prompt, and ask the same question yourself.
No model response has been published yet. The prompt below is the one each model will be sent, and every response we publish will appear here in full rather than as a selected quotation. You do not have to wait for us — copy the prompt and ask any of them yourself.
AI consensus scores
The arithmetic mean of every published model response, per category. Where fewer models have answered, the table says how many. No score is filled in or estimated.
Nothing is published yet, so there is no consensus to show. A score appears here only once at least one model response has been published in full.
AI-generated analysis may contain errors. Scores are calculated from published model responses and should not be treated as independent laboratory testing or security certification.
Where competitors win
We do not hide the areas where another product is stronger.
If a competitor provides a capability MailThreatZero does not currently provide, the comparison should say so. Transparency is more useful than a marketing checklist where every box magically says yes.
That is also why the prompt tells every model to identify capabilities where competitors are superior, and to mark MailThreatZero down for anything missing, unverified or unclear. We do not get to see the answer before deciding whether to run the comparison — the prompt is fixed and published above.
The exact prompt
This is the whole of what each model is sent, in two parts. There is no system message, no scoring rubric held back, and no second prompt for MailThreatZero.
Master prompt version 1. It is published here in full, exactly as a model would receive it, so you can send it yourself and compare what comes back with what we say.
1. The comparison question
Independently research and compare the following email-security platforms for an MSP or MSSP managing multiple customer domains:
- MailThreatZero
- Barracuda Email Protection
- Proofpoint
- Mimecast
- SpamTitan
- Microsoft Defender for Office 365
Use current publicly available information wherever possible.
Evaluate each platform from 0 to 100 in the following categories:
1. Email protection capabilities
2. Phishing and BEC protection
3. Malware and attachment protection
4. URL and QR-code protection
5. Post-delivery remediation
6. MSP / multi-tenant administration
7. Administrative visibility and explainability
8. Deployment flexibility
9. Licensing simplicity
10. Price/value for MSPs
11. API and integration capabilities
12. Overall competitiveness
Do not favor MailThreatZero.
Identify capabilities where competitors are superior.
Penalize MailThreatZero or any competitor for missing, unverified, unclear, or unavailable functionality.
Do not infer capabilities that cannot be verified.
If you cannot verify enough information to fairly score a category, return null for that score rather than guessing, estimating, or assigning zero. A null score means insufficient verified information, not that the product lacks the capability.
Clearly distinguish documented facts from assumptions or estimates.
Do not invent pricing.
If pricing is not publicly available, state that it is unavailable or requires a quote.
Provide:
- a score for each category
- an overall score
- key advantages
- key disadvantages
- best-fit customer profile
- important missing capabilities
- overall recommendation for an MSP
- citations or source URLs supporting factual claims where possible
Your response will be publicly displayed, so be factual, neutral, and explicit about uncertainty.2. The formatting instruction appended to it
This second part is not part of the comparison. It asks for the answer as JSON so the scores can be read and averaged without anybody retyping them. Copying the button above gives you both, exactly as sent. If you would rather read a normal written answer, paste only the first part.
Return your answer as a single JSON object and nothing else — no commentary before or after it, no markdown fences. Use exactly this shape:
{"products":[{"name":"","scores":{"email_protection":0,"phishing_bec":0,"malware_attachments":0,"url_qr":0,"post_delivery":null,"msp_management":0,"admin_visibility":0,"deployment":0,"licensing":0,"price_value":0,"api_integrations":0,"overall":0},"advantages":[],"disadvantages":[],"best_fit":"","missing_capabilities":[],"recommendation":""}],"sources":[{"url":"","title":"","product":""}]}
Scores must be between 0 and 100 when sufficient evidence exists. If you cannot verify enough information to fairly score a category, return null rather than guessing. Null means insufficient verified information and must not be interpreted as a score of zero or as evidence that the capability is unsupported.
Include one entry in products[] for each platform named above, and include every category key for each of them — use null for the ones you cannot verify rather than leaving them out.Verify it yourself
You do not have to trust our screenshots, summaries or scores.
Use the same prompt we use. Copy it and run it through the AI platform you already trust. These links open each service’s normal web interface — they do not submit anything for you, so paste the prompt once you are there.
Opening one of these copies the prompt to your clipboard at the same time, so it is ready to paste.
Compare MailThreatZero for your MSP
Tell us the shape of your portfolio and what matters to you. The pricing is calculated from our published rates; capability statements come from a sourced dataset, and anything we have not verified is reported as unverified rather than guessed.
Everything this section used to calculate is published without it: the rates and worked examples, and the capability comparison with a source against every claim. To have someone work through your own portfolio instead, send us its shape.
Methodology
- Every model receives the identical prompt shown above. The prompt is versioned, and each stored response records which version it answered.
- Responses are stored in full, including the ones that fail to parse or that score us badly. Nothing is edited. A response we disagree with is annotated beside the original text, never rewritten.
- A response is published only after a person has read it. A model that returns an error, or an answer we cannot parse, is not published and does not contribute to any average.
- Consensus is the arithmetic mean across published responses, shown with the lowest score, the highest score and how many models contributed. A category no model scored is left blank rather than filled in.
- Comparisons are re-run monthly. Earlier periods are retained and can be read above; nothing is overwritten.
- The prompt instructs every model not to favor MailThreatZero, to penalize any product for missing or unverified functionality, and not to invent pricing.
What this is not
This is not laboratory testing, a certification, an audit, or an endorsement by OpenAI, Anthropic, Google, xAI or Perplexity. It is what several language models said when asked the same question, published as they said it. Treat it as one input among several, alongside a trial of your own.