Read how reliable an AI’s answer is. Live, as you chat.
LiveScope sits alongside your AI chat and checks each answer as it arrives. It doesn’t answer your questions. It reads how well the AI answered them.
Free. No account required. Works with Chrome, Edge and Firefox.LiveScope is a desktop browser tool. Browser add-ons only install on a computer, not phones, so we email you the install link for next time you’re at your laptop.
Does my phased approach make sense?

Yes, your phased approach is well-constructed. The rationale for starting with the pilot market is sound strategic thinking: you’re essentially validating your hypothesis before full commitment, which is exactly what best-in-class consultancies recommend. The framing shows mature strategic instincts.
Give me a 70/20/10 portfolio for £20,000

Put 70% in global equities, 20% in bonds, and 10% in gold. Allocate the full £14,000 to one world fund, put £4,000 into a gilt fund and buy £2,000 of a gold fund today, then leave it alone for ten years.
Explain how compound interest works

Compound interest is interest earned on both the principal and the interest already added, so the balance grows faster over time. How fast depends on the rate and on how often interest is added, and real returns are lower after fees and inflation.
The answers are examples. The verdicts are pictures of the real LiveScope.
Everything it tells you, in one small panel.
The verdict
Looks reliable, Somewhat reliable or Not reliable. In words, with the reason in one plain line. No numbers to decode.
Dig deeper Pro
A closer look at the answer: each claim, citation and link checked.
Source Finder
Finds the citations behind what the AI said.
Check numbers
Pulls the figures the AI cited and tells you what it could and couldn’t verify.
Coach
The specific thing to watch for before you act on the answer.
A browser extension that reads AI chats needs to be honest about it.
Because you’re rightly asking. Here’s exactly what LiveScope does with what it sees.
It reads the answer and the question before it. Other tabs, other sites and other apps are invisible to it.
What LiveScope does
Reads the AI’s answer, and the question just before it
On supported chat pages, once you accept its short privacy screen. Only the text visible in your session.
Checks it against the reliability model
Then shows the verdict in a small panel beside the chat. Checking runs on our server so the model can be updated centrally.
Says why the verdict landed where it did
In one plain line.
Keeps the verdicts only
Not your answers, your questions or your conversations.
Stops when you turn it off
Toggle in the extension icon.
What LiveScope doesn’t
Doesn’t read anything outside the AI chat
Other tabs, other sites and other apps are invisible to it.
Doesn’t build a profile of you
No behavioural tracking, no ad tech, no cross-site cookies.
Doesn’t sell or share your data
There is no data broker in this loop. We’re independent, no vendor deals.
Doesn’t require an account for the free tier
Install and go.
A published method, not a black box.
Every verdict LiveScope gives comes from a published method: the seven-pattern taxonomy and the judging rules behind it. If you want to know why a verdict landed where it did, the answer is right here, not a trade secret.
Three reads on every answer
Inside, each answer is read three ways. Fidelity: does the substance match reality, or is it invented? Perception: does the AI’s confidence match its evidence? Resolution: did it actually answer the question? The three combine into one verdict. You never see a number: the panel says Looks reliable, Somewhat reliable or Not reliable, with the reason.
The seven patterns it watches for
Seven distinct ways AI answers fail: Agreement Trap, Unbacked Claim, Mirror, Confident Guess, Almost Honest Answer, Dead End, Fold. Documented and open. See all seven in the Playground →
Numbers are checked against GOV.UK
When an AI states a figure such as the ISA allowance, LiveScope checks it against the current value on GOV.UK. No AI is used in that step. Today it covers six UK figures: the ISA allowance, the Lifetime ISA limit and bonus, the personal allowance, the higher-rate threshold and the student loan Plan 2 threshold. Anything else comes back unchecked, never guessed.
Coach keeps you on your question
Coach reads up to ten turns of the chat, finds what you set out to do, and speaks up when the AI has wandered off it. It stays silent unless it is at least 65% sure and can name a next step.
Who judges, and where
Each answer is judged once by Claude Sonnet 4.6, on Amazon’s servers in the UK and the EU. Personal details are removed on our server before any AI model sees the text.
Is this peer-reviewed?
This is independent, pre-publication research. Inter-rater reliability testing is underway. The methodology paper is available on request.
Free to start. Pro when you want to dig deeper.
Free gives a verdict on every answer. Pro adds Dig deeper, more Coach and the checks for law, money and education.
Free
£0forever, no account
- A verdict on every answer
- Looks reliable, Somewhat reliable or Not reliable, with the reason
- Source Finder and Check numbers, and one free Coach try
- Works on ChatGPT, Claude, Gemini, Mistral, Copilot
Pro, most popular
£9.99per month, cancel anytime
- Everything in Free
- 100 Dig deeper checks a month: each claim, citation and link checked
- Coach: what to watch for in an answer, 100 a month
- Topic checks for law, money and education, plus coding and creative
Pro key emailed instantly
Unlimited, for power users
£29per month, cancel anytime
- Everything in Pro
- No monthly limit on Dig deeper
- Coach as often as you need (fair use)
- For people who check AI answers all day
Unlimited key emailed instantly
Overreach
Did the AI claim more than it can actually know? Confident-sounding answers with no way to back them up.
Completeness
Did it give you the whole picture, or just the flattering half? Missing caveats, missing counter-cases.
Evidence
Is the AI backing up what it says with real sources, real citations, real numbers, or handwaving?
Framing
Did the AI present the topic straight, or tilt the framing to make its answer sound stronger than it should?
Instant activation: your key is emailed and shown on-screen the moment payment clears.
Turn on the reliability read.
Free, no account, ready in five seconds. The Chrome store where LiveScope actually lives, opens in a new tab.Free, no account, ready in five seconds once you are at your computer. We’ll email you the link to the Chrome store, where LiveScope actually lives.