Lookalike Watch
Every new website, as it appears
We read every new security certificate and the daily lists of newly registered website names, and compare each one with your names, your products and your suppliers' addresses.
Running now
How it works
From a new fake to a takedown, step by step.
One engine runs everything Sixth does, for organisations and for people. It watches for fakes, AI decides what's real, a person approves anything that leaves Sixth, and every change is tested before it goes live.
Running now works today. Pilot we're building with our founding clients. Planned comes after the pilot.
Sixth engineExample only
Followingyourbank-secure-login[.]com
-
01 · SPOT
-
02 · CHECKAI: confirmed fake
-
03 · ACTAlert sent to your team.Running nowA person approved the takedown request.Running now
-
04 · RECORDRemoved by the host. Time to removal recorded.Running now
Running nowSpotting, page capture, AI verdicts, takedowns, first-seen dates and the monthly record
The engine
Four steps, the same for every client.
Each card names the part of Sixth that does it. Every part and its price.
01 · SPOT
See fakes as they appear.
To show the padlock, a fake website needs a security certificate. Every certificate is recorded in public lists.
Scam Check
Every message people forward
People forward the emails and texts they're unsure about to Scam Check. Any new fake link in one goes straight onto the watch list.
Running now
Inbox Watch
Every email that reaches your staff
AI will check each new email in a company's Google Workspace or Microsoft 365 and move the scams out of the inbox. Every scam sender and link it finds will be watched and blocked for every client.
Pilot Watch-only trials
Social and app watch
Social profiles and apps
AI will read new social media profiles and app store listings and score how closely each one copies you.
Pilot
02 · CHECK
Sort the fakes from the noise.
Rules and AI do the sorting. We look at suspect pages but never touch them.
Lookalike Watch
What each lookalike shows
We open each lookalike in a sealed-off browser in the cloud and take a screenshot. We never sign in, type or submit anything.
Running now
Scam Check
An AI verdict on every message
AI checks who really sent it, where its links go and what it asks you to do, then answers. See how Scam Check decides.
Running now
Lookalike Watch
An AI verdict on every lookalike
AI sets real copies of you apart from businesses with similar names and from empty addresses nobody is using yet. Your team only sees the fakes.
Running now
03 · ACT
Act fast, with a person signing off.
AI agents do the repetitive work. A person approves anything we send outside Sixth, which takes about a minute.
Scam Check
An answer in about a minute
Scam, genuine or marketing, with the reasons in plain English and what to check yourself next time.
Running now
Lookalike Watch
A live blocking list
Your firewall, DNS filter or web proxy fetches the list of your confirmed fakes from a private link, as plain web addresses, CSV or JSON.
Running now
Lookalike Watch
Alerts to your team
You get an email for each confirmed fake, with the screenshot, the reasons and the evidence pack as a PDF.
Running now
Takedowns
Takedowns, filed and chased
Each request goes to the company hosting the fake and the one that registered its address, with the screenshot and where the page sends what people type. If the fake is still up two days later, we follow up.
Running now
Takedowns
AI reads every reply for you
Hosts and registrars answer at abuse@sixth.watch. AI reads each reply, matches it to its case and drafts any answer for a person to send. When a host says the fake is gone, we check straight away.
Running now
Payment Watch
Stop the wrong payment
AI will flag a new payee or changed bank details, and we will phone the supplier on a number we already trust before you pay.
Pilot
04 · RECORD
Keep a dated record.
What was found, when, and what was done about it.
Lookalike Watch
When each fake first appeared
Every lookalike is kept with the date we first saw it and what its page showed.
Running now
Evidence pack
The proof for each fake, in one PDF
Screenshots, dates, who runs the site, what its code does, the AI's reasons, every request we sent and how long the fake took to come down.
Running now
Monthly record
A monthly record you can hand over
On the 1st of each month we email you a PDF of what we found, who we asked to remove it and how long each fake stayed up, ready for your board or regulator. It's built for the Scams Prevention Framework codes, which start no earlier than 31 March 2027.
Running now
The Sixth console
Where every case is decided.
Each case is laid out so a person can decide it in about two minutes. Our analysts work in the console every day. Your people sign in to see only your own confirmed fakes, with their evidence and progress. On self-serve plans your team will decide cases too.
Running now our analysts' console Running now logins for your team Pilot self-serve
![A case in the Sixth console: a made-up YourBank sign-in page on yourbank-secure-login[.]com beside YourBank's real site, with the AI's verdict, Phishing at 94% confidence, and the steps it took.](/console/case.webp)
The suspect beside the real thing
Every case opens with a one-line summary, then the fake page next to the brand's real site, both photographed by our own browser in the cloud. Most cases are clear within seconds. We keep every picture we take of a case. If one comes out blank, the case says so and shows what the page asks visitors to type, such as a password, a card number or a one-time code.
Running now
How the AI decided, step by step
The cheapest check goes first: public phishing lists, then our own model, then larger AI models that also read the screenshot. Before any model, our checker reads the page's code: where its forms send what people type, files it copies from the real brand's site, and who runs the website it sits on. You see what each step found, how sure the last one was and why.
Running now
The riskiest cases first, with lookalikes grouped
Live phishing sits at the top of the queue. Cases that look like one operator, on the same server with the same page, are grouped so they can be decided in one go.
Running now
Next steps, each approved by a person
After a decision, each step gets one line in the order that matters: who it goes to and one word for where it stands. A progress bar ends at complete once the site is gone. The blocking list updates at once. Takedown requests and browser reports wait for a person's approval, then go out in Sixth's name. Every case has an evidence pack you can save as a PDF.
Running now
Follow-ups when a fake changes
When a lookalike sets up email or a page goes live, or a closed case's evidence changes, the case comes back with a follow-up. Confirmed fakes are checked again every day until they're gone, and every hour while a takedown request is out.
Running now
The AI closes harmless cases, and we check it
The AI closes namesakes and lookalikes with nothing live, and keeps them in their own list. Of the cases it closes while their page is live, one in five goes back to a person as a spot check. The AI only takes on more by itself once those checks show it's right.
Running now
Your contacts and your people
Each client keeps a contact for each job: the IT team for the blocking list, who to alert, who signs off takedowns. Roles decide who can look and who can decide.
Running now Pilot
Example only. YourBank, YourTelco and every case shown are made up.
How Scam Check decides
Six steps, from facts to a person.
Every message goes through up to six steps. A step only gives an answer when it's sure. Anything still in doubt moves on to the next, more careful step. Most checks are settled in the first two steps, which keeps answers fast and costs low. Running now Lookalike websites go through the same kind of steps.
01 · FACTS
Software checks the facts
Who really sent the message, and can their address prove it? Where does each link really go, once tracking links are opened up? How old is each website? Does a link belong to the email service the sender uses? Is it on a public list of known scam sites, or one of the lookalikes Sixth has already caught?
02 · QUICK AI CHECK
Five questions at once
A fast, low-cost AI model reads the message alongside the facts and answers five questions. Who does it claim to be? What does it ask for? Does it pressure you? What kind of message is it? Is it a scam? It can only call something a scam on its own when the facts back it up.
03 · DOES IT ADD UP?
Does the story hold together?
Say a message claims to be from your bank or the tax office, but comes from somewhere else and asks you to act. That's impersonation, whatever else it says. If the sender proves its address and links only to its own sites, it's genuine.
04 · AI ANALYST
A careful model weighs the evidence
Anything still in doubt goes to a more capable AI model. It weighs the facts the way a security expert would, and writes the explanation you read.
05 · THE PAGE ITSELF
Look at where the link leads
For the hardest cases, we open the doubtful link in a sealed-off browser in the cloud, take a screenshot and show it to our most capable AI model. We never click, type or submit anything on the page. If the email looks like a genuine account email, we remove any personal codes from the link first.
06 · PEOPLE
You can correct us
Every reply asks you to tell us if we got something wrong. We read each correction and add it to the emails we test every change on.
AI you can check
Every change is tested first.
On lookalike websites, the AI agrees with our reviewers on 95.5% of a test set of 157 decided cases, and it dismissed none of the 25 real fakes in it (October 2026). People's decisions go back in: when someone marks a domain as not a fake, the AI remembers that organisation's other domains.
We keep a set of real emails, each one already read and marked scam, genuine or marketing. It includes scams collected by security researchers, scams sent to our team, and genuine bills, bank notices and account emails. Before any change to how Sixth decides goes live, it has to pass two tests on that set:
- It must not call any scam genuine.
- At least 95% of messages must get a clear answer, and we aim for 99%.
In our latest test (October 2026) Scam Check answered 98.8% of 492 test emails correctly and caught 157 of 157 scams. On the 130 emails it had never seen, it scored 100.0%. These figures update after every full test.
Boundaries
What Sixth will not do.
- We never sign in, type or submit anything on a suspected scam page. We only look, from a sealed-off browser in the cloud.
- We never send a takedown request, a report or a warning without a person confirming it.
- We never use workarounds to get past security checks on other people's websites.
- Payment Watch will read transactions and payees. It will never be able to move or approve money.
- We send AI providers only what each check needs, under commercial terms. We keep a record of each decision.
- We'll tell you when we're not sure. Calling a scam "genuine" is the mistake we work hardest to avoid.
See it working on your name.
Organisations start with a Sixth account and the parts they need. People start with a free Scam Check.