How we score chatting agencies
Chat Score rates chatting agencies on 15 criteria in four groups: price clarity (35%), creator control (25%), documented chat quality (25%) and proof (15%). Every cell needs a verbatim quote from the agency's own site, saved with a date. Anything unpublished scores zero. Price level is not scored; it lives in the cost calculator.
Updated 1 November 2026 · Methodology v1.0
What does Chat Score measure?
Chat Score measures how much a chatting agency puts in writing about the things that protect a creator: price and basis, control of login and payouts, how chat quality is ensured, and proof. It does not measure sales skill, which no outsider can verify. It covers 43 reviewed agencies; it is not a census of every agency in the market. Included are chatting agencies and full-management agencies that publicly offer chatting or fan messaging on their own site; the inclusion rule is the public offer, not the score. A published full-management rate is a bundle price, so it scores as “rate only” (6) and is not used as a chat-only price in the calculator or the price statistics.
Which criteria and weights are used?
Fifteen criteria in four groups. The group score is the mean of its criteria; the total is the weighted sum: A price clarity 35%, B creator control 25%, C documented chat quality 25%, D proof 15%. Rounded to one decimal. Ties break on the chat price cell, then login, then alphabetically.
| Group | Criterion | Bands (fact → points) | With evidence |
|---|---|---|---|
| A | Chat-only rate published | 10 · rate and basis stated 6 · rate or range only 2 · only 'from' or 'custom' 0 · nothing published | 32/43 |
| A | What the fee is charged on | 10 · clear basis with a worked example 7 · clear basis, no example 3 · 'revenue' without a basis 0 · not stated | 26/43 |
| A | No extra fees | 10 · no setup/hourly/minimum fee stated 6 · extra fee disclosed 0 · not stated | 32/43 |
| A | What the price includes | 10 · itemised list (chat, PPV pricing, mass messages, customs, report) 5 · partly listed 0 · not stated | 37/43 |
| B | Login and 2FA stay with the creator | 10 · sole access stated 5 · shared access disclosed 0 · not stated | 29/43 |
| B | Payouts go directly to the creator | 10 · stated 0 · not stated | 23/43 |
| B | Term and notice | 10 · no minimum term and notice of 30 days or less 6 · no minimum term 3 · fixed term with an exit route 0 · not stated | 35/43 |
| B | No commission after the end | 10 · no tail commission stated 5 · tail commission disclosed 0 · not stated | 12/43 |
| C | Who writes the messages | 10 · humans only, no AI chat product 6 · humans with disclosed AI assistance 3 · AI chatting sold 0 · not stated | 20/43 |
| C | The creator's voice | 10 · onboarding/persona brief and an approval or limits rule 5 · one of the two 0 · not stated | 27/43 |
| C | Coverage and languages | 10 · 24/7 and at least two chat languages 6 · 24/7 or two languages 0 · not stated | 38/43 |
| C | Reporting | 10 · cadence and a public sample report 6 · cadence stated 0 · not stated | 24/43 |
| D | Company behind it | 10 · name, jurisdiction, address and register 5 · name and country 0 · not stated | 22/43 |
| D | Results | 10 · figures with artefact and a way to verify 6 · figures with artefact 3 · figures only 0 · none | 27/43 |
| D | Who it is (not) for | 10 · minimum revenue or situations, plus an alternative 5 · one of the two 0 · not stated | 20/43 |
What counts as evidence?
Only text on the agency's own website. We save each page (HTML and text, with SHA-256 and capture date); pages that need JavaScript are captured with a browser and marked as rendered. Every scored cell carries at least one verbatim quote, and a check script fails the build if any quote is not found in the saved snapshot. A fact found only in a site's llms.txt scores half the band. Calls, emails and promises do not count.
Why is the price level not scored?
Because a lower percentage is not evidence of better chatting, and because the basis changes the bill more than the number. Price level is compared separately in the cost calculator. Agencies are scored on whether they publish a price and its basis, not on how high it is.
Is any score set by hand?
No. A person selects the facts and quotes for each cell; fixed band rules turn facts into points; code computes groups, totals and rank. The same rules apply to every agency, including White Door Management, which is connected to this site. The dataset and every cell are on the data page.
How do corrections work?
Send the page that states a fact to editorial@chattingagencies.com. We capture it, check the quote, re-run the score for every listed agency and log the change with the date. Colour thresholds are the same for everyone: green from 7, yellow from 4, grey below.