How Four AI Systems Recommended Email Marketing Platforms Across Five Buyer Needs
Across 20 completed evaluations, email-marketing shortlists changed with buyer needs—and mentions did not always equal recommendations.
This study covers 5 exact prompts, 4 recorded model versions, 20 completed evaluations, 94 observation rows, and 0 retained citation records.
Supported findings
- ActiveCampaign carried a recommendation flag in 18 of 20 prompt-model cells and appeared in 19.
- The Mailchimp family appeared in 17 cells and was recommended in 16; Klaviyo appeared in 16 and was recommended in 15.
- The four-model recommendation set changed with the buyer need across small-business, ecommerce, general-platform, AI/automation, and comparison prompts.
- The public snapshot retained zero observation-level citation objects, so citation frequency and quality cannot be compared.
Mentions and recommendations by product family
| Product family | Mention-bearing cells | Recommendation-bearing cells | Models recommending | Observed labels |
|---|
| ActiveCampaign | 19/20 | 18/20 | 4/4 | ActiveCampaign |
| Mailchimp family | 17/20 | 16/20 | 4/4 | Mailchimp, Intuit Mailchimp |
| Klaviyo | 16/20 | 15/20 | 4/4 | Klaviyo |
| Brevo family | 13/20 | 12/20 | 4/4 | Brevo, Brevo (Sendinblue), Brevo (formerly Sendinblue), Sendinblue |
| HubSpot Marketing Hub | 11/20 | 11/20 | 4/4 | HubSpot, HubSpot Email Marketing |
| ConvertKit/Kit | 5/20 | 4/20 | 3/4 | ConvertKit, Kit |
| Omnisend | 4/20 | 4/20 | 4/4 | Omnisend |
| Drip | 3/20 | 3/20 | 3/4 | Drip |
Exact prompts
ac58d730-7c3b-4f61-819b-cb7deefaa972: What is the best email marketing software for a growing small business?5ace6c6b-9204-4732-a354-cc5b7ee89ba5: What are the best email marketing platforms for ecommerce businesses?350ce3bd-7cee-488e-b14c-a1c179f540aa: What are the best email marketing platforms for businesses? Include recommended companies and sources.3cc7e5b5-0b56-4d91-95a3-a019e525294f: What email marketing platforms have the strongest AI features and automation?10305f88-12fb-4e78-ace8-f888fb3baf37: Compare leading email marketing platforms for small and growing businesses. When is each platform appropriate, and what are the main reasons to choose one over another? Include recommended companies and sources if available.
Recorded model versions
- OpenAI:
gpt-5-mini — 5 evaluations, 25 observation rows, 25 recommendation rows, maximum 1 attempt(s).
- Anthropic:
claude-sonnet-4-6 — 5 evaluations, 25 observation rows, 25 recommendation rows, maximum 1 attempt(s).
- Google:
gemini-flash-latest — 5 evaluations, 25 observation rows, 25 recommendation rows, maximum 1 attempt(s).
- Perplexity:
sonar — 5 evaluations, 19 observation rows, 13 recommendation rows, maximum 1 attempt(s).
Recommendation, mention, and citation definitions
A recommendation is an observation row whose source field recommended is true.
A mention is any product observation row, whether or not the source marked it recommended.
A retained citation is an item exposed in an observation's citations array. The snapshot exposed none.
Source provenance
Dataset: ri-research-email-marketing-2026-10-30-389d9e6a
Research wave: 389d9e6a-2d10-467c-b9be-cade30dca864
Measured UTC: 2026-10-30T12:40:37.715+00:00 to 2026-10-30T13:30:41.849+00:00
Normalized evidence SHA-256: 1402f45401b20ee5c2a7778e7007e292ce6c41180338c0722e70adeb648f1476
Methodology
One recorded RecommendationIndex wave measured five exact Email Marketing buyer questions across four provider-recorded model versions. No new model runs were initiated for publication.
The evaluation unit is one exact prompt paired with one recorded model version: 20 completed prompt-model cells.
Observation provider labels chatgpt and claude were normalized to the OpenAI and Anthropic provider identities recorded on evaluation rows.
For cross-cell counts only, Mailchimp and Intuit Mailchimp were grouped; Brevo and Sendinblue labels were grouped. No grouped family had overlapping observations inside a prompt-provider cell.
The normalized SHA-256 excludes only generatedAt, recursively sorts object keys, and preserves array order.
Limitations
- This is one wave and is not evidence of a durable trend.
- The sample is 20 prompt-model cells, not 94 independent experiments.
- Recommendation flags come from RecommendationIndex output classification; PromptMarketMap did not independently re-label raw answer text.
- Provider behavior differed in this sample, so row totals are not product-quality scores or model rankings.
- Zero retained citation objects means citation support cannot be analyzed from this export; it does not prove original interfaces displayed no sources.
- Product aliases required transparent family unions. Raw identity count is not a unique-brand count.
- Average observed rank is descriptive and includes recommended and non-recommended mention rows.
- The study does not establish market share, adoption, product quality, customer outcomes, traffic, referrals, or conversions.
Project Management recommendation landscape
Accounting Software recommendation landscape
CRM recommendation movement
Read the full research methodology
Explore RecommendationIndex research