ObsurfableShort answer: Data-infrastructure buyer prompts do not produce one shared vendor leaderboard across AI engines. Search-native engines (Gemini, Perplexity, Claude) treat Bright Data as the default #1 in this category — 32.8% of Perplexity answers and 24.4% of Gemini answers lead with Bright Data, with mention rates of 72.7% and 65.2% respectively. ChatGPT leads with Snowflake on 8.8% of category prompts and mentions Bright Data on only 9.5%. On 73 matched ChatGPT–Perplexity prompts, Perplexity still puts Bright Data first 34.2% of the time versus 21.9% for ChatGPT, and mentions Bright Data on 72.6% of identical questions versus 45.2% — while Snowflake essentially disappears from Perplexity’s shortlist (0% as #1, 0% mentions on the matched panel).
The split is not random ranking noise. It is different house favorites on the same buyer questions.
What we analysed
We used Obsurfable Explorer’s research corpus: buyer-style prompts tagged Technology · Data Infrastructure — web scraping, proxies, datasets, pipelines, warehouses, and adjacent data-collection tooling.
For each prompt we keep the latest observation per platform family (ChatGPT, Gemini, Claude, Perplexity, Grok), preferring web-app captures over API runs when both exist. We report:
- #1 pick — first brand extracted from the answer
- Mention rate — brand appears anywhere in the extracted list
We built matched panels where the same prompt was observed on two engines within the category window.
| Scope | Value |
|---|---|
| Corpus window | 30 June – 20 October 2026 |
| Data Infrastructure prompts | 1,334 |
| Total observations (category) | 3,032 |
| ChatGPT + Perplexity matched panel | 73 |
| ChatGPT + Gemini matched panel | 83 |
Browse live category views on Data Infrastructure, ChatGPT, and Perplexity. See Bright Data’s corpus profile for mention patterns across engines.
What we found
1. Search-native engines cluster on Bright Data; ChatGPT clusters on warehouses and orchestration
Category-wide #1 picks (latest observation per prompt per engine):
| Engine | Prompts observed | Bright Data #1 | Snowflake #1 | No brand extracted |
|---|---|---|---|---|
| Perplexity | 128 | 32.8% | 0.0% | — |
| Gemini | 201 | 24.4% | 0.5% | — |
| Claude | 128 | 19.5% | 0.0% | — |
| ChatGPT | 1,289 | 5.6% | 8.8% | 28.0% |
Mention rates tell the same story with larger gaps:
| Engine | Bright Data mention | Snowflake mention |
|---|---|---|
| Perplexity | 72.7% | 3.1% |
| Claude | 68.0% | 1.6% |
| Gemini | 65.2% | 7.0% |
| ChatGPT | 9.5% | 21.1% |
Among Gemini + Perplexity answers combined, Bright Data is the #1 pick on 26.1% of prompts — more than triple ChatGPT’s 5.6% Bright Data #1 rate. ChatGPT’s top #1 brands instead include Snowflake (8.8%), Fivetran (3.3%), and Airflow (2.7%) — a warehouse/ETL-flavored shortlist that search-native engines rarely lead with.
2. Matched prompts: Perplexity and Gemini still elevate Bright Data on identical questions
ChatGPT + Perplexity matched panel (n=73)
| Metric | ChatGPT | Perplexity |
|---|---|---|
| Bright Data #1 | 21.9% (16/73) | 34.2% (25/73) |
| Bright Data mention | 45.2% | 72.6% |
| Snowflake mention | 1.4% | 0.0% |
| Different #1 vs paired engine | — | 75.3% of prompts |
ChatGPT + Gemini matched panel (n=83)
| Metric | ChatGPT | Gemini |
|---|---|---|
| Bright Data #1 | 8.4% (7/83) | 14.5% (12/83) |
| Bright Data mention | 28.9% | 61.4% |
| Cloudflare #1 | 0.0% | 12.0% |
| Different #1 vs paired engine | — | 77.1% of prompts |
Scraping-adjacent vendors widen with retrieval-heavy engines on the same prompts:
| Brand (mention rate) | ChatGPT (n=73) | Perplexity (n=73) | Gap |
|---|---|---|---|
| Bright Data | 45.2% | 72.6% | +27.4 pp |
| Apify | 37.0% | 49.3% | +12.3 pp |
| ScrapingBee | 16.4% | 28.8% | +12.4 pp |
| Oxylabs | 26.0% | 31.5% | +5.5 pp |
Snowflake and Databricks do not appear in this matched scraping-heavy overlap — the disagreement is which data-collection stack the engine opens with, not a one-rank swap on a shared list.
3. Live web search tracks the Bright Data skew
On the 73-prompt ChatGPT–Perplexity panel, retrieval usage diverges:
| Engine | Web search active | Avg cited sources per answer |
|---|---|---|
| ChatGPT | 24.7% | 3.60 |
| Perplexity | 93.2% | 10.71 |
On the 83-prompt ChatGPT–Gemini panel, Gemini runs web search on 78.3% of answers versus 24.1% for ChatGPT, with 5.24 average sources versus 3.37.
When engines retrieve the open web on data-infrastructure questions, answers disproportionately surface proxy and scraping vendors (Bright Data, Apify, Oxylabs) — categories that appear far less often as ChatGPT’s default #1 when retrieval is inactive.
4. Claude confirms the search-native cluster
Claude’s category sample (128 prompts) aligns with Gemini and Perplexity, not ChatGPT: 68.0% Bright Data mention rate and 19.5% Bright Data #1. Snowflake appears as #1 on 0.0% of Claude answers in this window.
That makes the split three independent search-native families versus ChatGPT, not a single outlier engine.
Why this is surprising
“Data infrastructure” sounds like a warehouse and pipeline market — Snowflake, Databricks, Fivetran. Buyer prompts in this tag also include scraping, proxies, and dataset access, where a different vendor set competes.
Yet even controlling for prompt identity on matched panels, Perplexity names Bright Data first on one-third of the same questions where ChatGPT does not — and three in four matched prompts produce a different #1 brand across engines. A visibility program that tracks ChatGPT alone will under-count Bright Data and over-weight Snowflake relative to what Perplexity, Gemini, and Claude show the same buyers.
This is the same structural pattern we documented for transactional email APIs — different engines, different house favorites — but here the fault line runs between retrieval-heavy scraping shortlists and ChatGPT’s warehouse-orchestration defaults, not between two API incumbents.
Limitations and caveats
- Matched panels are small (73–83 prompts) because only a subset of category prompts have been observed on both ChatGPT and a search-native engine. Overlap skews toward scraping and collection use cases where Perplexity and Gemini are more likely to run.
- ChatGPT dominates sample volume (1,289 prompts with observations vs 128–201 for other engines). Category-wide ChatGPT rates are more stable; engine-specific rates for Gemini, Perplexity, and Claude will move as coverage grows.
- Brand extraction depends on answer text; 28.0% of ChatGPT answers in this category yield no extracted brand #1, which depresses apparent concentration.
- Results describe recommendation text in this corpus, not market share, revenue, or product quality.
Methodology
Platforms: ChatGPT (web UI and API captures mapped to one family), Gemini, Claude, Perplexity, and Grok where sample size allows. API-only captures are deprioritized when a web UI observation exists for the same prompt and family.
Prompt set: Buyer prompts in Technology · Data Infrastructure in Explorer’s public research corpus.
Deduplication: Latest observed_at per prompt × platform family; platform keys mapped into families as listed above.
Metrics: #1 brand (first extracted mention), mention rate (any position), web-search flag when reported by the capture, and source URL counts per answer.
Exclusions: Observations tied to private website-scoped prompt sets are omitted from this research slice.
Reproduce individual prompts at explorer.obsurfable.com. Signed-in users can view observation history per prompt.
Conclusion
On data-infrastructure buyer prompts in Obsurfable’s corpus, search-native engines default to a Bright Data–centric scraping shortlist — 65–73% mention rates and 20–33% #1 rates — while ChatGPT defaults to a warehouse-oriented leaderboard led by Snowflake (8.8% #1) with Bright Data at 5.6%. On 73 matched ChatGPT–Perplexity prompts, Bright Data mention rates diverge by 27 percentage points on identical questions, and Snowflake vanishes from Perplexity’s list.
Measuring “AI visibility” in this category without per-engine recommendation panels will blend incompatible house favorites into a single ranking that no buyer actually sees.
Browse the free Explorer corpus, or match evidence to your company in the Obsurfable App.