Short answer: In transactional email API buyer prompts, when an engine links a vendor’s own domain, it usually names that vendor in the recommendation list — unlike developer-media outlets, where mentions and links diverge. Among 473 deduplicated UI observations with at least one cited URL in Obsurfable’s API Platforms slice, Resend is mentioned on 100% of answers that cite resend.com versus 46.6% when that domain is absent (+53.4 pp). Postmark shows the same pattern (98.2% vs 48.1%, +50.2 pp). The exception is parent-brand confusion: six answers cite sendgrid.com but zero name SendGrid — five name Twilio instead, and nine of 37 answers that cite twilio.com omit any Twilio-family brand from the list.
Citation-based visibility tracking and mention-based tracking largely agree in this category — until Twilio/SendGrid or other nested brands break the mapping.
What we analysed
We used Obsurfable Explorer’s research corpus: buyer-style prompts tagged Technology · API Platforms — transactional email, webhooks, deliverability, and developer send APIs.
For each prompt we keep the latest observation per platform family (ChatGPT, Gemini, Claude, Perplexity, Grok), preferring consumer chat captures over API batch runs when both exist. We conditioned mention rate on whether the answer’s source list included the vendor’s primary domain (e.g. resend.com, postmarkapp.com, mailgun.com, brevo.com, twilio.com, sendgrid.com).
| Scope | Value |
|---|---|
| Corpus window (UI rows in category) | 6 August – 3 November 2026 |
| API Platforms prompts (UI subset) | 232 |
| Latest UI-family observations | 577 |
| Observations with ≥1 cited source | 473 (82.0%) |
| Raw category observations (all captures) | 7,901 |
Primary vendors: Resend, Postmark, Mailgun, Brevo, Twilio, SendGrid — names as extracted from answer lists; domains matched from cited URL hostnames.
Browse live aggregates in API Platforms on Explorer and engine splits such as ChatGPT and Perplexity.
What we found
1. Own-domain citations strongly predict vendor mentions
On the 473-observation sourced panel (latest per engine per prompt):
| Vendor | Answers citing vendor domain | Mention rate when cited | Mention rate when not cited | Lift |
|---|---|---|---|---|
| Resend | 78 | 100.0% | 46.6% | +53.4 pp |
| Postmark | 57 | 98.2% | 48.1% | +50.2 pp |
| Mailgun | 40 | 87.5% | 32.8% | +54.7 pp |
| Brevo | 25 | 68.0% | 12.1% | +55.9 pp |
| Twilio | 37 | 75.7% | 41.1% | +34.6 pp |
164 answers cite at least one of these tracked vendor domains; 157 (95.7%) also name at least one matching vendor in the extracted list. Retrieval is doing most of the recommendation work when sources appear.
Brevo shows the widest absolute gap between cited and uncited mention rates (+55.9 pp) — the vendor is rarely named from training alone (12.1% when brevo.com is absent) but appears on 68.0% of answers once the domain is linked.
2. SendGrid links attach to Twilio, not SendGrid
SendGrid breaks the pattern:
| Signal | Value |
|---|---|
Answers citing sendgrid.com (any subdomain) | 6 |
| Also name SendGrid in the brand list | 0 (0%) |
| Name Twilio (not SendGrid) | 5 (83.3%) |
Example behavior on ChatGPT-sourced rows: answers link sendgrid.com or ahoy.sendgrid.com while the ordered list leads with Resend or Twilio — the cited product documentation does not flow into the extracted recommendation string.
Across 37 answers citing twilio.com, 9 (24.3%) omit any Twilio-family brand from the list despite linking Twilio-owned URLs.
3. ChatGPT shows the largest citation-conditioned lift
Among platforms with enough cited rows, ChatGPT shows the steepest lift when a vendor domain appears:
| Vendor | ChatGPT: mention when domain cited | ChatGPT: mention when not cited | Lift |
|---|---|---|---|
| Resend | 100% (n=24) | 35% (n=37) | +65 pp |
| Postmark | 96% (n=28) | 33% (n=33) | +63 pp |
| Mailgun | 92% (n=13) | 21% (n=48) | +71 pp |
| Brevo | 100% (n=4) | 7% (n=57) | +93 pp |
Grok and Perplexity show the same direction on Resend and Postmark (typically +39 to +61 pp on cited vs uncited splits) but with more sourced answers per prompt — see category platform pages for live mention leaders.
4. Contrast with developer media
In the same research corpus, developer-media sourced answers decouple mentions from links: Hashnode is named on 65% of cited answers but linked on 5%. API Platform vendor domains couple the signals: if postmarkapp.com is in the source list, Postmark almost always appears in the recommendation bullets.
That difference is category-specific, not engine-universal.
Why this is surprising
Email API buyers often assume “if we are not cited, we are invisible.” In this slice, absence of a vendor domain is the stronger negative signal — mention rates fall to the 30–50% range for major vendors when their site is not linked, even though those vendors still appear often from training-style lists.
The SendGrid/Twilio split is equally important for measurement: citation share for sendgrid.com will not move SendGrid’s mention share if engines credit parent Twilio or competitors instead — a failure mode that mention-only dashboards would miss entirely.
Limitations
- 82% of latest UI observations in this subset include sources — higher than developer media (~39%), but 18% still have no URLs to condition on.
- Domain matching uses hostname rules; regional docs, redirects, or third-party blog posts may cite vendors without matching our primary-domain list.
- Brand extraction depends on list order and canonical strings; “Twilio SendGrid” variants are counted separately from Twilio and SendGrid.
- Sample covers 232 prompts in API Platforms — not the full multi-vertical corpus.
- API batch captures are deprioritized when a UI capture exists; results reflect consumer chat surfaces.
Methodology
Platforms: ChatGPT, Gemini, Claude, Perplexity, and Grok UI captures in the public research corpus.
Prompts: Active buyer-style API Platform queries in the shared Explorer research set (no customer website scope).
Deduplication: Latest observation per prompt per platform family; UI platform keys preferred over API aliases.
Mentions: Canonical brand strings in the extracted ordered list; #1 = first named brand.
Citations: Hostnames parsed from each observation’s source URLs; vendor “cited” when the primary commercial domain substring matches (e.g. postmarkapp.com for Postmark).
Lift: Percentage-point difference in mention rate between observations that cite the vendor domain and those that do not, within the same sourced panel.
Exclusions: Non-UI platform keys; observations outside the API Platforms subvertical.
Date window: Earliest to latest UI observation in this category in the November 2026 corpus export (2026-08-06 – 2026-11-03).
Conclusion
For transactional email APIs, vendor-domain citations are a high-precision signal for who gets named — Resend, Postmark, and Mailgun almost always appear when their docs are linked, and fall to roughly half (or less) of answers when they are not. Programs that track only link share or only mention share will mostly agree here, except where Twilio/SendGrid (and similar nested brands) decouple the hostname from the extracted recommendation. Score mentions conditioned on whether your domain was cited, not either metric alone.