Frequently asked questions
Last updated 17 September 2026
What's the difference between Quick Scan and Deep Verify?
Quick Scan gives you a trust score in about fifteen seconds — bias, purpose, funding and credibility, scored by three AI models at once. It costs 1 credit.
Deep Verify goes further: it pulls out the specific factual claims in an article and checks each one against a real source — a government dataset, a court record, an official report — rather than just judging the article's overall tone. It takes longer (closer to a minute) and costs more credits, since it's doing meaningfully more work per claim.
How do credits work?
Every account gets a monthly credit allowance depending on your plan — free accounts get a small amount to try the product, paid plans get more. Credits reset each billing cycle and don't roll over. A Quick Scan and a Deep Verify cost different amounts, shown right on the button before you run either one, so there's never a surprise.
If an analysis fails for any reason — a page that can't be reached, a site blocking automated access, or an error on our end — you are not charged a credit for that attempt. You're only ever charged once you actually get a result back.
Why did fetching a URL fail?
Pasting a URL works on most sites, but some — particularly sites with strong anti-bot protection, or ones that load their article text via JavaScript in a way our fetcher can't always see — will fail to load properly. This is a genuine limitation of fetching pages automatically, not something wrong with your account.
The reliable fallback: copy the article text yourself and paste it directly into the box. That always works, since it skips the fetching step entirely. If you're using the Chrome extension rather than the website, this issue doesn't come up at all — the extension reads the page directly from your browser rather than fetching it separately.
How do I manage or cancel my subscription?
Click the account icon (top right) and choose Manage subscription — this takes you straight to a secure Stripe page where you can change plans, update your payment method, or cancel outright. There's no need to email us to cancel; it's fully self-service.
I forgot my password — how do I reset it?
On the sign-in screen, click Forgot password? and enter your email. You'll get a link by email to set a new password. If you don't see it within a few minutes, check your spam folder.
What do the scores actually mean?
The overall score blends factual accuracy, source credibility, transparency and funding — averaged across three independent AI models (Claude, GPT-4 and Gemini). 70 and above is generally trustworthy, 40–69 means read with some caution, and below 40 is high risk.
Purpose is different from the other scores — it describes what a piece is trying to do (inform vs persuade), not whether it's good or bad. A persuasive piece isn't automatically untrustworthy; it's just a different kind of writing than a neutral news report.
Why do the three AI models sometimes disagree?
That's intentional, and it's actually useful information. Refract shows you when Claude, GPT-4 and Gemini land on similar scores (high consensus, more confidence in the result) versus when they're genuinely split (a sign the story is more contested or ambiguous than a single score alone would suggest).
Is my data safe? Do you store what I read?
We don't store the full text of articles you analyse. Article text is sent to AI providers to generate your result and isn't retained by us afterwards. See our privacy policy for the full detail.
Something's not working — how do I get help?
Email hello@readrefract.com and we'll get back to you. Including the article URL or a screenshot of what you saw helps us sort it out faster.