As of 13 August 2026, AI can only partly run an A/B test on your marketing emails.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open your email platform and record its A/B testing options, audience exclusions, consent settings, tracking controls and reporting fields.
   2. Gather the current email, approved brand guidance, offer details, landing-page URL, audience definition and any past campaign results you can verify.
   3. Paste those materials into the prompt and ask the model for one hypothesis, two single-variable versions and the setup checklist.
   4. Compare both drafted emails against your approved offer, price list, landing page, unsubscribe process and brand guidance, then correct unsupported claims.
   5. Configure the test in your email platform with matched audience rules, exclusions, tracking links, one fixed test variable and a documented primary metric.
   6. Send a test email to internal addresses and check links, rendering, personalisation, unsubscribe controls and tracking before obtaining campaign approval.
   7. Run the test through your platform, export the results into the results table, and ask the model to compare the versions without inventing missing figures.
   8. Check the model's interpretation against the platform report and your campaign objective before choosing a winner or sending a follow-up email.

   Prompt

   Help me plan and prepare an A/B test for a marketing email. Do not claim to access, configure or send anything in my email platform.
   
   Business and campaign context:
   - Business: [business and offer]
   - Audience: [who will receive it]
   - Existing email or draft: [paste email]
   - Campaign objective: [the single primary objective]
   - Email platform: [platform name]
   - Brand and legal constraints: [tone, claims, exclusions and consent requirements]
   - Available audience size and relevant past results: [paste only figures I can verify]
   - Test window and practical limits: [details]
   
   Produce:
   1. One clear test hypothesis.
   2. Two versions that change only one meaningful variable, such as the subject line, preview text or call to action. Keep all other content and targeting the same.
   3. A test setup checklist for my email platform, including audience splitting, exclusions, send timing, tracking links, suppression rules and final-send controls.
   4. A primary success metric and sensible secondary checks. Do not invent benchmarks, sample sizes or statistical results.
   5. A results table with fields for version, recipients, deliveries, bounces, opens if available, clicks, conversions, unsubscribes and complaints.
   6. A plain-English method for comparing the results, including what would make the result inconclusive. Do not declare a winner from incomplete data.
   7. A list of every factual, legal, consent or platform-specific item I must check myself before sending.
   
   Write concise copy in British English. Preserve factual claims from my source material, flag anything unsupported, and do not invent customer details, performance figures or compliance approval.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.