Questions, answered plainly

No, and the difference is the whole product. A chatbot gives you a number it generated. We give you a number arithmetic produced, and show the arithmetic.

Ask a chatbot to score an income claim and it will produce a confident, plausible figure. Ask it twice and you may get two different figures. It cannot tell you where the number came from, because there is no “where” — it was generated, not computed.

Our pipeline uses a language model for exactly one narrow job: reading. It extracts what the claim literally states — revenue, timeframe, units, tools — at temperature 0 with a fixed seed, and anything the claim doesn’t state stays null rather than being guessed. The model never scores.

Everything after that is code: fixed formulas over version-stamped price tables, platform payout latency, and market price bands, plus an internal-consistency check and documented hype penalties. Same inputs produce the same score every time, and every teardown publishes the calculation so you can recompute it yourself.

Because a score you can’t inspect is just an opinion with a number attached — which is exactly what we exist to push back on. We can’t demand receipts from other people and then refuse to show our own.

Ratings organisations work this way for the same reason: test protocols and rating criteria are published, because publication is what makes a verdict auditable, challengeable, and worth trusting.

The formula was never the hard part. The hard part is the maintained price tables, the growing corpus of scored claims and its history, the probe work behind each replication playbook, and shipping it every single Friday. Those take sustained labour, not a copy-paste.

Then we want to hear it, and the score changes. Every teardown publishes its inputs and its arithmetic precisely so an error is findable. If a price table is stale, a claim stated something we missed, or the math is off, send it to challenge@realityscore.tech.

Scores carry the model version and the price-table date they were computed against, so a re-score is transparent rather than a silent edit.

No. A score measures whether a claim is feasible as stated — not whether the claimant is honest. Most low scores come from claims that simply don’t state enough to be checked.

In fact the most common outcome in our corpus isn’t a debunk — it’s INCOMPLETE: the claim omits the timeframe or the unit count, so no honest score is possible. We publish that rather than guessing the missing pieces.

The full leaderboard of scored claims, every teardown with its complete workings, the published methodology, the weekly claim-intelligence findings, and a few self-serve checks so you can run a claim yourself. No card required.

The Friday Audit Brief with the week’s verdicts and the deeper intelligence write-up, plus the replication playbook — the priced bill of materials, unit economics, and go/no-go gates behind a specific play — and a larger monthly allowance of self-serve checks.

Plans and current pricing are on the plans section.

Paste a link (X and YouTube work directly) or the claim text. We pull the post, extract the stated facts, show you what we extracted so you can correct it, then score it instantly and show every calculation.

That correction step matters: extraction is the one non-deterministic part of the pipeline, so we surface it rather than hiding it. Results are labelled an instant estimate to distinguish them from the curated weekly teardowns. Try it here.

Cancel any time — it stops future renewals and you keep access through the period you’ve already paid for. If you’re unhappy, email support@realityscore.tech within 14 days of a charge and we’ll refund that period in full, no questions asked. Full terms are in the Terms.

Yes — send it here. The strongest submissions are public posts that state real numbers, because those can actually be checked. Claims are selected for the weekly cycle partly on how widely they’ve spread.

No. RealityScore is published for informational and educational purposes only. It is not financial, investment, or business advice, and nothing here promises or guarantees any earnings. Scoring a claim as feasible is not a recommendation to pursue it. See the Disclaimer.

We’d rather state these than have you discover them:

• We score what the claim states, using public evidence. We cannot see anyone’s bank account.
• Prices move. Every score is stamped with the price-table date it used.
• Extraction is the one model-driven step and can misread an ambiguous post — which is why self-serve checks let you correct it before scoring.
• A feasible score means the arithmetic works, not that you will get that result.
• Our corpus is still small; sample sizes are published beside every statistic on the insights page, and we don’t draw trends the sample can’t support.

RealityScore™ is published by Machine Digital Inc. The work is deliberately judged on its arithmetic rather than on anyone’s personal brand — every verdict is reproducible from published inputs, so it stands or falls on the math. Reach us at support@realityscore.tech.