How Trust Cue rates accounts

What we read

We read a sample of an account's recent public posts on X — what they have posted and re-posted publicly. That sample is the evidence for the rating, so a rating describes an account's recent public posting, not its whole history. Trust Cue may also open a public link from the account's profile to check an identity or credential claim; that check alone can never raise the quality of the rating. Trust Cue does not access your private session, read DMs, or post for you.

Every account runs through the same five checks. The checks are automated; no one decides a rating by hand, and the rules don't bend for any account.

How language coverage works

Trust Cue reads posts in the language they were written in. It never translates a post into English and then scores the translation — too much of what makes a claim careful or careless dies in translation. Accounts that post in more than one language are sampled across each language they use substantially, so a smaller but meaningful part of their posting cannot be averaged away.

We only rate a language once it has cleared the same quality bar we hold English to, end to end. If an account posts substantially in a language we cannot yet read at that standard, Trust Cue does not guess — it returns no rating, and it does not charge you for one. If the gap is narrower, such as a single profile claim we cannot read confidently, we drop that one conclusion and say so, and the rest of the rating stands.

The rating is the same in every language. Translation changes the wording you read — never the mark, the confidence, the evidence, or the original quotes, which always stay available exactly as posted.

The five checks

Each rating is built from five checks. The check names are stable — they mean the same thing on every account so you can compare one to another. What changes per account is the evidence behind each one.

1. Subject depth

   Does the account actually know the topic they post about — real methodological or domain substance — or are they performing expertise? We weigh the depth of the substance, not the volume of the posts.

2. Original framing

   Do they add their own framing, or mostly boost and re-post others? Amplification is fine, but an account that only forwards other people's work reads differently from one that reasons in its own voice.

3. Sticks to the facts

   Do the posts show sourcing, honest uncertainty, corrections — or contradictions? An account that hedges where it should and links its evidence scores higher than one that overstates. This watches for factual discipline; it does not independently verify every claim.

4. Upfront about their stake

   Are they upfront about what they have at stake? Selling a course in your domain isn't low-credibility — undisclosed promotion is. This check flags shill and pump patterns, not the mere existence of a product.

5. Who engages with them

   Who engages with them, and who do they amplify? We look at the accounts they actually interact with, not just names they drop. When we can't see enough of that, we treat it as a gap in our evidence rather than a flaw in the account.

Evidence or it doesn't count

Every scored check is bound to a word-for-word quote from a real public post. If we can't point at the post that supports a signal, the signal isn't claimed. A quote proves the account said something, not that the statement is true — so each quote is weighed against everything else we read, and a single flattering line cannot carry a check the wider sample doesn't support.

Why a high rating is hard to earn

The top rating — "Strong support" — means the posts we read showed strong support under all five checks. It is deliberately hard to reach:

• Thinner evidence can't earn the top mark. When we can only read an account weakly, the best it can reach is "Mixed picture". The strongest mark has to be earned on strong evidence, so a thin read can never produce one.
• A hidden stake caps the rating. If an account quietly sells what it posts about, the rating is capped at "Thin support" no matter how good the content is. Strong writing can't read as trustworthy when the author has an undisclosed interest in what you believe.
• "Too little to tell" is not "Thin support". A private or newly-joined account with too little public posting reads as low-information, not as untrustworthy. We refuse to rate rather than guess — and we always tell you which of the two it is.

When we can't rate, we say so

If there isn't enough public posting to rate an account fairly, Trust Cue says "Too little to tell" honestly instead of inventing a rating. That is a real result, not a failure: knowing the record is too thin to read is often the useful answer, and it is reached by the same five checks as any other rating.

What ratings are not

A Trust Cue rating is an automated, account-level signal from a sample of recent public posts. It is not an endorsement, a character judgment, a comprehensive fact-check, or financial advice. "Strong support" means the posts we read scored strongly under the stated checks; it does not mean every claim is true or that you should act on it.

What the rating can miss

Trust Cue can miss deleted posts, anything said outside the posts we sampled, private conduct, new information published after the rating was made, or a conflict of interest that was never disclosed publicly. Automated reading and interpretation can also simply be wrong. Evidence links let you inspect the basis of a signal, but they do not remove the need to verify an important claim against primary sources.

The result is designed for repeatable source triage: deciding where to look more closely. It is not a substitute for reporting, due diligence, or expert review when a decision carries material consequences.

The 48-hour freshness rule

A rating reflects the posts we read at the moment it was made. A rating you run stays valid for 48 hours, after which it expires — because an account can post a lot in two days, and a stale signal is worse than an honest "check it again." Re-checking reads the account fresh. This is also why a shared link never serves a stale rating: share pages only show ratings that are still fresh and were made under the current rules.

The sample ratings on the welcome page are the deliberate exception. They are dated snapshots of real accounts, kept fixed so everyone sees the same worked examples — the extension labels them as frozen samples and shows you when each one was captured. They are there to show you what a rating looks like, not to tell you what those accounts are posting today.

Same rules for everyone

There is no per-account tuning, no override, and no human in the loop. Every account gets the five checks, the evidence rule, and the caps above — including our own accounts. If the rules ever change, the change is documented here on this page, and every rating made under the old rules stops being served, so no one ever reads an old rating under new rules. The dated sample snapshots described above are the one exception: they stay readable as fixed examples of how a rating reads, and they always carry the date they were captured on.

Common questions

How do I use Trust Cue on X?

Trust Cue adds a small badge beside names in your X feed. Tap a badge and that account's rating opens in a side panel — the overall mark, the five checks, the public posts quoted behind each one, and the caveats. Nothing runs until you tap.

What does Trust Cue read?

A sample of an account's recent public posts on X. Trust Cue may also open a public link from that account's profile to check an identity or credential claim. It never reads your private session or DMs, and never posts for you.

How does Trust Cue rate an X account?

Every account runs through the same five automated checks — subject depth, original framing, sticks to the facts, upfront about their stake, and who engages with them — each backed by a word-for-word quote from a real public post. No one decides a rating by hand.

Does Trust Cue fact-check every claim or endorse an account?

No. Trust Cue reads patterns across an account's recent public posting. It does not independently verify every claim, endorse a person, or replace checking an important claim against primary sources.

What happens when there isn't enough to rate an account?

Trust Cue says "Too little to tell" instead of guessing. That is a real result reached by the same five checks as any other rating — knowing the public record is too thin to read is often the useful answer.

How fresh are Trust Cue ratings?

A rating you run is read from recent public posts and stays fresh for 48 hours. After that Trust Cue re-reads the account rather than showing a stale signal.

Are the sample ratings live?

No, and they are labelled that way. The samples are dated snapshots of real accounts, kept fixed so every new install sees the same worked examples. They show you what a rating looks like, not what those accounts are posting today.

What does Trust Cue cost?

Adding the extension is free and the sample ratings are free to read. Checking any other account starts at $3, or $18/month for 20 ratings.

Is a Trust Cue rating an endorsement or financial advice?

No. Ratings are automated signals from an account's recent public posts — not endorsements, and not financial advice.