Sample report

What ChatGPT & Claude say about ProductClank.

2
Campaigns
24
Submissions
24
Transcripts read
2
Models · ChatGPT + Claude
The consensus verdict

Both models call ProductClank “a real platform” and “a clever execution of a real trend.” They love the structural bet — performance-only rewards, creators as partners. But independently, almost every transcript lands on the same two doubts: the amplification mechanics flirt with platform ToS, and the growth story is “good vibes, thin evidence.” Bottom line, in Claude’s words: promising directionally — but they’d want to see verifiable outcomes before committing budget. Which is exactly what the last quarter of shipping has been about — every flag below now carries what we did about it.

How the models describe it

The description the AI reaches for, unprompted.

Community-led growth platformPay-for-performance, not ad spendCreators as partners, not ad channelsBoost — coordinated amplificationDiscovery — AI replies from real accountsContent-creator campaignsX · LinkedIn · Reddit · Instagram · TikTok · Farcaster10 replies / 30 likes per boostStars / social-capital system“No upfront ad spend, no equity”12,000+ users · 1,000+ campaigns
What they consistently praise

The strengths that land.

Performance-only rewards align incentives

“Creators earn when milestones are hit, not just for posting” — repeatedly called a smart structural choice that “aligns incentives better than flat-fee influencer deals.”

Both models

Creators framed as partners, not ad channels

Flagged as “a meaningful differentiation from traditional influencer marketing platforms.” The framing itself is the moat.

Claude + ChatGPT

“No upfront spend, no equity” fits the audience

“Genuinely well-suited to cash-poor, time-rich Web3 and early-stage startups. Cheaper and faster than traditional influencer deals.”

Both models

A real flywheel, plus Stars

Seen as “the actual mechanics of a flywheel, not just messaging around one.” The Stars layer read as “a smart nod that contribution should feel valued beyond cash.”

Claude
What they consistently flag

The risks — ranked by how often they came up.

These are the objections an AI hands a buyer who asks about you first. We're not hiding them — and where we've shipped a fix, it's right there under the flag.

  1. 1

    Coordinated amplification looks like a ToS risk

    Likes / retweets / bookmarks “on demand” were flagged in nearly every response as sitting “uncomfortably close to engagement manipulation,” carrying “real platform-policy risk on X.”

    Most-cited — nearly universal
    What we didShipped

    Every reply now passes a relevance gate — off-topic and low-effort submissions are rejected before they ever post — and rewards are tied to genuine, on-topic advocacy, not likes-on-demand. It's belief-driven distribution with guardrails, not a bot farm.

  2. 2

    Thin, unverifiable evidence

    “No independent third-party reviews (G2, Trustpilot) or named case studies — most of what's public is the company's own site and blog posts.” The “95–98% cost savings” figure was called out directly as “marketing copy, not audited data.”

    Both models, repeatedly
    What we didShipped

    We opened a third-party review presence (Trustpilot) so the evidence isn't only self-published, and swapped the unaudited “95–98%” headline for a sourced range with a stated method. Named case studies with real outcomes are rolling out next.

  3. 3

    Incentives can erode the authenticity CLG runs on

    “Community-led growth works best when the community genuinely believes in the product. Incentivized content at scale can erode authenticity fast — the thing that makes CLG valuable in the first place.”

    Claude
    What we didShipped

    AI-vision proof review now samples submissions for genuine engagement, and a 3-strike system with reward clawback removes farmers. Quality is enforced, not assumed — the incentive rewards real advocacy, not volume.

  4. 4

    Still early-stage; network effects unproven

    “At ~12K users and $100K distributed, it's a small, early ecosystem — network effects, the lifeblood of a marketplace like this, are unproven at scale.”

    Claude
  5. 5

    Outcomes depend on who runs the campaign

    “Niche builder tools with genuinely passionate early users will get far more out of it than generic SaaS products trying to manufacture buzz.” Also noted as X-centric — may complement rather than replace Farcaster-native community building.

    Claude
In their words

Verbatim.

Unedited pulls from the submitted transcripts — the exact language a prospect's AI produces.

Claude · asked by our communitypraised
The performance-only model is a smart structural choice that aligns incentives better than flat-fee influencer deals.
Claude · asked by our communityflagged
The “95–98% cost savings” claim is marketing copy, not audited data — good vibes, thin evidence.
ChatGPT · asked by our communityflagged
The ProductClank MCP is currently provided as a remote MCP/connector for Claude, not for ChatGPT.
ChatGPT · asked by our communitypraised
Users earn points for completing simple, public, repeatable actions that spread the project's name naturally.
Notable findings

What else the transcripts surfaced.

The GoodDollar / ChatGPT run exposed an MCP onboarding wall — and we're rebuilding it

The task asked people to install the ProductClank MCP in ChatGPT, but the connector was Claude-only — so ChatGPT users hit a mismatch and several logged “first try, failure.” We're not glossing over it: the MCP onboarding wasn't good enough, and it's being rebuilt now with a ChatGPT-compatible path, clearer prerequisites, and a copy-paste config. Worth noting the idea itself landed — once engaged, ChatGPT rebuilt a full ProductClank-style campaign unprompted. The concept works; the connect step is what we're fixing.

What to do about it

Turn each objection into a fix.

MoveAnswersPriorityStatus
Reframe “coordinated amplification”
Move away from likes/retweets “on demand” toward genuine-belief advocacy; surface the relevance gates and proof-review guardrails already running.
ToS + authenticityHighShipped
Substantiate or retire “95–98%”
Replaced the unaudited headline with a sourced range and stated method — an AI will quote it either way.
“Marketing copy, not data”HighShipped
Publish verifiable proof
A third-party review presence (Trustpilot) is live so evidence isn't only self-published; named case studies with real outcomes are rolling out.
“Thin evidence,” “hard to verify”HighIn progress
Fix the MCP onboarding mismatch
The connector was Claude-only, yet the GoodDollar task asked people to install it in ChatGPT — so they failed. Rebuilding a ChatGPT-compatible path with stated prerequisites and a copy-paste config.
GoodDollar run — “first try, failure”HighIn progress
Report quality, not just volume
Publish content-retention and review pass rates. Claude said this is exactly where these models “prove out or fall apart.”
Authenticity at scaleMedIn progress
Keep feeding structured, factual content
AEO is working — both models described you accurately. Seed third-party sources so the AI's skeptical section shrinks over time.
Reinforces the winMedShipped

Sources — two Take-Action campaigns: “ProductClank” (Ask Claude, public network) and “Ask ChatGPT about ProductClank” (GoodDollar space). 24 submissions; all AI transcripts read from shared links and screenshots. Quotes transcribed verbatim. Sample skews toward reward-motivated participants on one prompt phrasing — treat frequencies as directional; the convergence across independent submissions is the signal.

Want this report for your product?

Real people ask your buying questions across ChatGPT & Claude and send back the answers.

Run a check