ObsurfableShort answer: When the same transactional-email API prompts are run through ChatGPT and Perplexity, the engines do not just disagree on ranking — they recommend from different vendor pools. On 117 matched prompts in Obsurfable's API Platforms category, Perplexity puts Notify first 47% of the time and mentions it on 66% of answers. ChatGPT puts Notify first only 28% of the time — and crowns Twilio on 27% of the same questions. Perplexity almost never leads with Twilio (1%). The pattern repeats on a 192-prompt ChatGPT–Grok panel: Grok names Notify first 38% of the time versus 17% for ChatGPT.
This is platform bias, not buyer-intent noise. The prompt is identical; the house favorite changes with the engine.
What we analysed
We used Obsurfable Explorer's research corpus: buyer-style prompts tagged Technology · API Platforms — questions about transactional email APIs, webhooks, delivery logs, and developer-focused send infrastructure (not marketing automation suites).
For each prompt we keep the latest observation per platform family (ChatGPT, Perplexity, Grok, Gemini, Claude), preferring web-app captures over API runs when both exist. We then built matched panels: prompts observed on two or more engines within the category window.
| Scope | Value |
|---|---|
| Corpus window | 17 July – 23 September 2026 |
| API Platforms prompts | 197 |
| Total observations (category) | 725 |
| ChatGPT + Perplexity matched panel | 117 |
| ChatGPT + Grok matched panel | 192 |
Primary metrics: #1 pick (first brand named in the answer) and mention rate (brand appears anywhere in the extracted list) for nine transactional-email API vendors: Notify, Postmark, Resend, Twilio, Mailgun, Amazon SES, Mailtrap, Brevo, and SendGrid.
Browse live examples in API Platforms on Explorer and Notify's corpus profile.
What we found
1. Perplexity and Grok share a Notify-first shortlist; ChatGPT defaults to Twilio
On the 117-prompt ChatGPT–Perplexity panel, #1 picks diverge sharply:
| #1 brand named | ChatGPT | Perplexity |
|---|---|---|
| Notify | 28.2% (33/117) | 47.0% (55/117) |
| Twilio | 26.5% (31/117) | 0.9% (1/117) |
| Postmark | 15.4% (18/117) | 16.2% (19/117) |
| Resend | 4.3% (5/117) | 12.8% (15/117) |
| No brand extracted | 9.4% (11/117) | 6.0% (7/117) |
On 19.7% of matched prompts (23/117), Perplexity's #1 is Notify while ChatGPT's #1 is a different brand — usually Twilio, Postmark, or Mailgun.
The Grok panel confirms the search-native cluster. On 192 matched ChatGPT–Grok prompts:
| #1 brand named | ChatGPT | Grok |
|---|---|---|
| Notify | 17.2% (33/192) | 37.5% (72/192) |
| Twilio | 15.6% (30/192) | 3.6% (7/192) |
| Postmark | 7.8% (15/192) | 3.6% (7/192) |
| Resend | 2.6% (5/192) | 9.4% (18/192) |
Grok's category-level top brands already skew Notify-first (API Platforms on Grok); ChatGPT's category view leads with Twilio (API Platforms on ChatGPT).
2. Mention rates show the same split — not just ordering
#1 position can flip on a single rank change. Mention rates show the underlying vendor preference more clearly:
ChatGPT + Perplexity matched panel (n=117)
| Brand | ChatGPT mention rate | Perplexity mention rate | Gap |
|---|---|---|---|
| Notify | 35.0% | 65.8% | +30.8 pp |
| Resend | 38.5% | 52.1% | +13.6 pp |
| Twilio | 57.3% | 42.7% | −14.6 pp |
| Mailgun | 55.6% | 41.0% | −14.6 pp |
| Postmark | 53.0% | 51.3% | −1.7 pp |
| Amazon SES | 44.4% | 35.9% | −8.5 pp |
Notify moves from a minority mention on ChatGPT to a supermajority on Perplexity — on the same questions. Twilio and Mailgun invert: incumbents ChatGPT surfaces more often are down-ranked when Perplexity answers the identical prompt.
ChatGPT + Grok matched panel (n=192)
| Brand | ChatGPT | Grok | Gap |
|---|---|---|---|
| Notify | 21.4% | 43.8% | +22.4 pp |
| Resend | 21.9% | 34.9% | +13.0 pp |
| Twilio | 32.8% | 24.0% | −8.8 pp |
| Mailgun | 31.8% | 22.9% | −8.9 pp |
3. The split tracks live web search, not category ignorance
Search-native engines in this panel run live retrieval far more often:
| Engine | Matched panel | Web search active | Avg sources per answer |
|---|---|---|---|
| ChatGPT | n=117 (vs Perplexity) | 3.4% | 0.08 |
| Perplexity | n=117 | 67.5% | 9.20 |
| ChatGPT | n=192 (vs Grok) | 2.1% | 0.10 |
| Grok | n=192 | 100% | 7.96 |
When Perplexity or Grok retrieves the web, Notify and Resend — newer developer-focused transactional email APIs — appear in the retrieved shortlist more often. ChatGPT's API captures in this corpus rarely attach sources and more often reproduce a training-era incumbent stack (Twilio, Mailgun, Postmark).
That does not make either list "wrong." It means platform choice is vendor choice in this category: measuring visibility on ChatGPT alone will miss the Notify-heavy shortlist Perplexity and Grok ship to the same buyers.
4. Category context: the disagreement is concentrated, not universal
Across the full API Platforms category (197 prompts), Postmark, Notify, and Resend cluster at the top in aggregate — but platform × category views diverge:
| Platform | Prompts observed | Top 3 brands (category aggregate) |
|---|---|---|
| ChatGPT | 197 | Twilio, Mailgun, Postmark |
| Perplexity | 117 | Notify, Postmark, Resend |
| Grok | 192 | Notify, Postmark, Resend |
The matched-panel tables above control for prompt identity. The aggregate split is not driven by Perplexity seeing a different question set — on overlapping prompts, the house favorites differ.
Why this is surprising
Transactional email APIs look like a mature, interchangeable market: Twilio SendGrid, Mailgun, Postmark, Amazon SES, and a wave of developer-first challengers (Resend, Notify). Buyer prompts in this category rarely name a preferred vendor.
Yet on identical questions:
- Perplexity is 1.9× more likely than ChatGPT to put Notify first (47% vs 28%).
- ChatGPT is 29× more likely than Perplexity to put Twilio first (27% vs 1%).
- Grok mentions Notify on 2× as many matched prompts as ChatGPT (44% vs 21%).
A brand tracking only ChatGPT would read Twilio and Mailgun as co-leaders. The same brand on Perplexity would see Notify as the default recommendation on nearly half of matched prompts. That is not a ranking shuffle — it is a different recommendation market per engine.
Limitations and caveats
- Category scope: Findings apply to Technology · API Platforms prompts in Obsurfable's research corpus (197 prompts). They do not generalise to all email, SMS, or marketing-automation categories.
- Panel size: The ChatGPT–Perplexity panel is 117 prompts; Grok corroboration uses 192. Gemini and Claude have too few overlapping observations in this category (two three-way matches) to quote reliably.
- ChatGPT capture mix: Most ChatGPT observations are API runs without attached citations. UI captures cite more often, but the matched panel already holds the question constant — the vendor bias persists.
- Brand extraction: Brands are extracted from response text. Notify (notify.cx) is a real transactional-email API in the corpus; we do not treat generic verbs as brands.
- Temporal window: Observations span 17 July – 23 September 2026. Vendor momentum (e.g. Resend's growth) may shift these rates over time.
Methodology
Corpus: Obsurfable Explorer research prompts — publicly browsable buyer questions with repeated AI observations. We included active, non–website-scoped research prompts tagged Technology · API Platforms.
Platforms: ChatGPT (web app + OpenAI API family), Perplexity, Grok (web app + xAI API family). Gemini and Claude are reported only where matched sample size exceeds 50 prompts.
Matching: A prompt enters a panel when it has a latest observation on each engine in the pair. ChatGPT–Perplexity: 117 prompts. ChatGPT–Grok: 192 prompts.
Brand counting: We use the ordered brands_mentioned list from each observation. #1 pick = first brand. Mention rate = share of panel prompts where the brand appears anywhere in the list. Email-API vendors are counted individually (Notify ≠ generic "notify" verbs).
Source and search flags: sources = URLs attached to the answer. web_search_used = engine reported live search for that observation.
Exclusions: Website-scoped customer prompts are excluded from the research corpus. Observations outside the category tag are not included in category panels.
Conclusion
In transactional email APIs, AI engines have house favorites. Perplexity and Grok systematically overweight Notify and Resend on the same buyer prompts where ChatGPT leads with Twilio and Mailgun. On 117 matched prompts, Perplexity puts Notify first 47% of the time; ChatGPT does so 28% of the time while putting Twilio first 27%. Mention-rate gaps are even wider: Notify appears on 66% of Perplexity answers vs 35% on ChatGPT.
If you sell or compete in developer infrastructure, treat each engine as a separate recommendation market. A single-score "AI visibility" metric will average away exactly the platform bias buyers experience when they switch from ChatGPT to Perplexity.
Explore the data: API Platforms category · Perplexity in API Platforms · Notify