We asked ChatGPT, Perplexity and Gemini fifty questions a real Webflow buyer asks — by service, industry, geography and company size. They named 376 providers. Exactly two are known to all three engines.
50 Webflow buyer questions · 150 answers · July 16, 2026 · US / English · brand names come from answer text — all three engines count
| # | Provider | Mentions | Engines | |
|---|---|---|---|---|
| 1 | Finsweet | 90 | 3 / 3 | all-three club |
| 2 | Flow Ninja | 62 | 3 / 3 | all-three club |
| 3 | Edgar Allan | 39 | 2 / 3 | |
| 4 | Refokus | 39 | 2 / 3 | |
| 5 | Flowbase | 20 | 1 / 3 | templates — not an agency |
| 6 | Relume | 18 | 1 / 3 | components — not an agency |
| 7 | Veza Digital | 14 | 1 / 3 | |
| 8 | Flowout | 12 | 2 / 3 | |
| 9 | BX Studio | 10 | 1 / 3 | |
| 9 | 8020 | 10 | 2 / 3 | |
| 9 | Amply | 10 | 1 / 3 | |
| 9 | Shadow Digital | 10 | 1 / 3 |
Finsweet's lead is enormous — 90 mentions is more than positions 5 through 12 combined. And note rows 5 and 6: asked for an agency, the engines put two products in the top six. They don't reliably distinguish "who can build this for me" from "what tool exists."
Our general-market study found 79% of brands were single-engine. We expected a tight niche to consolidate — a small market where the engines converge on one short list. The opposite: 97%. Each engine has assembled its own, almost entirely private, map of the Webflow market.
If your prospect asks ChatGPT and their colleague asks Perplexity, they hear about different companies.
Unless you're one of two names, at most one of those conversations includes you.
At least two engines agreed on 80% of niche questions — but full three-way consensus halves when the questions get specialized. Niche depth doesn't create consensus; it removes the universally famous names that produce it. The "three separate games" finding is at its most extreme exactly where specialized businesses live.
We track 1,464 real Webflow-focused agencies (a proprietary roster built from partner directories and public rosters, matched by normalized name and domain root). Across all 150 answers to all 50 questions, 61 of them were named even once, by any engine.
There is a head — five names take 30% of all mentions — but the defining feature of this market's AI layer is absence: 95.8% of the real market is invisible to it. The gap between position 12 on the Index (10 mentions) and the median tracked agency (zero mentions, zero engines) is the entire game.
multi250.csv, Set C.Before optimizing anything. Ask all three your own buyers' questions — given the 97% finding, one engine's answer tells you nothing about the other two.
Finsweet and Flow Ninja publish prolific, structured, genuinely useful material on their own domains — the exact page profile this series keeps finding in citations. That's a content posture, not a budget.
Flowbase and Relume rank because engines conflate products with providers. Competing with component libraries for "best agency" mentions is the wrong battle.
Report No. 5 measured fortnight-scale churn in these same answers. A snapshot's losers are next month's entrants — we'll re-run it.
| Metric | Scope | Value |
|---|---|---|
| Questions / answers | Set C, all 3 engines | 50 / 150 |
| Named providers (after filter) | all 3 engines | 376 |
| Total mentions (after filter) | all 3 engines | 821 |
| Generic/marketplace entries removed | filter rule | 24 entries / 91 mentions |
| Single-engine providers | all 3 engines | 365 / 376 = 97% |
| Two-engine providers | all 3 engines | 9 |
| Known to all three | all 3 engines | 2 |
| Top-5 share of mentions | all 3 engines | 250 / 821 = 30% |
| Niche all-3 agreement | Set C | 15 / 50 = 30% |
| Niche ≥2 agreement | Set C | 40 / 50 = 80% |
| General all-3 agreement, same day | Set A | 59% |
| Tracked agencies ever named | internal roster, n=1,464 | 61 = 4.2% |
Only two providers were named by all three engines across 50 Webflow buyer questions: Finsweet (90 mentions) and Flow Ninja (62). 97% of the 376 named providers are known to just one engine.
61 of the 1,464 agencies we track (4.2%) were named even once by any engine — and that figure is a floor. For most of the market this is an existence problem, not a ranking problem.
Less. All-three agreement was 30% on the Webflow questions versus 59% on general questions asked the same day — specialization removes the famous names that create consensus.
Same method, pointed at your agency. We'll ask all three engines your buyers' questions and show you which ones name you — and which name someone else.
Run the free audit →