Our WorkAll ServicesAI VisibilitySEO AutomationProduct ManagementClaude AI ServicesZatrovo BookingStart a Project
content-automation · 8 min read · 2 November 2026

Spot Phishing Emails with Claude AI: A Check for UK Owners

A practical UK guide to spotting phishing emails in your business, with a copy-and-paste Claude prompt, the red flags to check and where to report scams.

Jacob Horgan, Founder, Irvale Studio
Jacob Horgan
Founder, Irvale Studio
UK small business owner checking a suspicious supplier invoice email on a laptop in a Leeds office.

Most phishing emails that reach a UK small business don't look like the clumsy "Nigerian prince" messages people still joke about. They look like a DPD redelivery notice, a Microsoft 365 password expiry warning, or an invoice from a supplier you really do use, sent on a Friday afternoon when you're trying to clear the inbox before the weekend. This guide shows you how to spot phishing emails in a business setting, gives you a Claude prompt you can paste in when something feels off, and is honest about where an AI check helps and where it doesn't.

How big is the phishing problem for UK small businesses?

Phishing is the attack UK businesses are most likely to face. The government's Cyber Security Breaches Survey 2025 found that 43% of businesses identified a breach or attack in the past year, and phishing was the most common type, hitting 85% of those businesses.

The figures come from the Cyber Security Breaches Survey 2025, published by the Department for Science, Innovation and Technology. The survey has ranked phishing as the most common type of breach or attack year after year, well ahead of viruses, ransomware or account takeovers.

43%UK businesses that identified a breach or attack in the last 12 months
Source: Cyber Security Breaches Survey 2025
85%Of those businesses, the share that experienced phishing
Source: Cyber Security Breaches Survey 2025

Those numbers only count attacks people noticed. The ones that slipped through unnoticed don't appear anywhere, which is why a habit of checking matters more than a one-off training session.

What are the red flags in a business phishing email?

Look for a sender address that doesn't match the company, urgency or secrecy, a request to change payment details or log in, link text that doesn't match where the link goes, and attachments you weren't expecting. One red flag is a reason to pause. Two or more means treat it as hostile.

The signs that matter most in a business inbox:

  • The display name and the address disagree. "HMRC Refunds" sent from a Gmail address, or "Sarah at Brakes" from a domain with an extra letter in it.
  • Pressure on the clock. "Account suspended in 24 hours", "final notice", "the director needs this before 3pm".
  • Changed bank details. A supplier "moving banks" is the classic invoice fraud setup. Real suppliers rarely announce this by email alone.
  • A login page you didn't ask for. Microsoft, Google, Xero and your bank don't need you to "verify" your password through a link in an unexpected email.
  • Links that hide their destination. Hover over the link on a desktop, or press and hold on a phone, and read the actual domain before you click.
  • Unexpected attachments, especially zipped files, HTML files or documents that ask you to "enable content".

Spelling mistakes used to be a reliable tell. Attackers now use AI writing tools too, so a polished email proves nothing either way.

How can Claude help you check a suspicious email?

Paste the email text, and the headers if you can get them, into Claude with a structured prompt. Claude goes through the red flags one by one, points out anything inconsistent and tells you what to verify. It's a fast second opinion that makes you slow down, not a verdict.

The value of using Claude here is less about clever detection and more about forcing a pause. Phishing works because people react quickly. Copying a message into a separate tool and reading a structured answer takes about a minute, and that minute is often enough to break the spell.

To get the headers in Outlook, open the message, go to the three-dot menu and choose the option to view the message source or details. In Gmail, use the three-dot menu and "Show original". Headers show the real sending server and whether checks such as SPF and DKIM passed, which is far more telling than the display name.

What prompt should you paste into Claude?

Use a prompt that asks Claude to list red flags, rate its confidence, explain its reasoning and say what you should check offline. Asking for reasoning stops you from trusting a bare "this looks fine".

Copy this, then paste the email underneath it:

You are helping a UK small business owner decide whether an email is phishing. Review the email and headers below. 1. List every red flag you find, quoting the exact text. 2. Compare the display name, From address, Reply-To address and any link domains, and say whether they match the claimed sender. 3. If headers are included, say whether SPF, DKIM and DMARC appear to pass. 4. Give a rating of likely safe, suspicious or likely phishing, with your reasons. 5. Tell me exactly what to verify by phone or in my own records before I act. Don't assume the email is genuine because it is well written.

Strip out customer names, account numbers and anything else personal before you paste. More on that below.

Where does an AI phishing check fall short?

Claude only sees what you paste. It can't open links safely, look up how old a domain is, see your supplier's real bank details or detect a genuine mailbox that has been hacked. Use it to catch patterns and prompt verification, never as permission to pay.

The honest trade-offs:

  • Compromised real accounts. If a supplier's actual email account is hacked, the sender address, headers and writing style can all look correct. Only an offline check catches this.
  • No live lookups. Claude can't confirm whether "paypa1-secure.co" was registered yesterday. It can point out that the domain looks wrong, but you still need judgement.
  • False comfort. A "likely safe" rating on a payment request still means you should verify. Make that a written rule.
  • Volume. Running every email through a chat window doesn't scale. If your inbox is the real bottleneck, a proper triage setup is a separate job, covered in this guide to AI email sorting for business owners.

Is it safe to paste business emails into Claude?

Generally yes, if you remove personal data first and check your privacy settings. On Free, Pro and Max plans you choose whether chats train Anthropic's models. Team, Enterprise and API use fall under separate commercial terms.

According to Anthropic's update to its consumer terms, announced in August 2025, Free, Pro and Max users choose whether their chats are used to improve Claude. If you allow it, data retention extends to five years. If you don't, the existing 30-day period applies. Claude for Work (Team and Enterprise) and the API are excluded from that change.

For a business, the practical rule is simple. Delete customer names, addresses, account numbers and anything you'd have to disclose under UK GDPR before pasting. The phishing signals sit in the sender details, links and wording, not in the personal data. For a fuller list of what to check before staff use AI tools, see this AI data protection checklist for UK businesses.

What should you do after spotting a phishing email?

Don't click anything. Forward it to the NCSC at report@phishing.gov.uk, delete it, and warn anyone else in the business who might have received it. If money or passwords are involved, call your bank and change the password immediately.

The National Cyber Security Centre's reporting page asks you to forward suspicious emails to report@phishing.gov.uk and says you can forward as many as you like. The NCSC uses these reports to take down the scam websites behind them. You don't need to forward anything your provider has already sent to spam or junk. Suspicious text messages can be forwarded free to 7726.

If someone has already clicked:

  1. Change the password for that account straight away, and anywhere else it was reused.
  2. Turn on two-step verification if it isn't already on.
  3. If bank details were entered or a payment was made, ring your bank on the number on the back of your card, then report the fraud to the police.
  4. Check the mailbox for new forwarding rules, a common trick attackers use to keep reading your mail.

How do you turn this into a habit for your whole team?

Write three rules on one page: never pay or log in from an email alone, always verify bank detail changes by phone using a number you already hold, and forward anything suspicious to the NCSC. Then make it safe for staff to pause and ask.

The rules only work if nobody gets told off for checking. Attackers often pose as the owner or a senior client precisely because staff feel awkward questioning them. Say out loud that a 30-second phone call is always acceptable, even if it turns out the email was real.

Businesses that handle client money face the highest stakes. Accountants, solicitors and conveyancers are frequent targets of invoice and completion-fund fraud. For them, verifying payment changes by phone should be written into the process, not left to individual judgement.

Is a Claude check worth adding to your routine?

For most UK small businesses, yes, as a pause button for the handful of emails that feel off each week. It costs a minute, catches obvious patterns and reminds you to verify. It doesn't replace phone checks, two-step verification or reporting to the NCSC.

Keep the prompt saved somewhere easy to reach, such as a pinned note or a Claude Project, so using it takes no effort. Pair it with the three written rules and two-step verification on every email account. That combination covers the attack the Cyber Security Breaches Survey says you're most likely to face, without buying a security product you'll never fully set up.

Next stepSet up Claude properly for your businessPrompts, privacy settings and workflows built around how your team actually works.
Common Questions

Spot Phishing Emails with Claude AI — FAQ

How common are phishing attacks on UK small businesses?

Very common. The UK government's Cyber Security Breaches Survey 2025 found that 43% of businesses identified a breach or attack in the previous 12 months. Phishing was the most common type, experienced by 85% of those businesses. In other words, phishing is the attack a typical owner is most likely to face. Bigger, more technical attacks happen far less often. That's why a simple, repeatable check on suspicious emails is worth more to most small firms than an expensive security product they never configure properly. The survey also only counts attacks people noticed, so the real figure is likely to be higher.

Can Claude tell me for certain whether an email is a phish?

No. Claude can only read the text and headers you paste in. It can't check whether a domain was registered last week, it can't see your supplier's real bank details and it can't confirm who sent the message. It's good at spotting the patterns phishing relies on, such as false urgency, mismatched sender addresses, odd link text and requests to change payment details. Treat its answer as a second opinion that helps you slow down. For anything involving money, passwords or bank details, confirm with the person or company by phone, using a number you already hold rather than one printed in the email.

Is it safe to paste a suspicious email into Claude?

Usually, yes, as long as you strip out personal data first. Remove customer names, account numbers and anything covered by your UK GDPR duties before pasting. Anthropic updated its consumer terms in August 2025 so that Free, Pro and Max users choose whether their chats are used to train models. If you allow training, data can be kept for up to five years. If you don't, the standard 30-day retention period applies. Team and Enterprise plans and the API fall under commercial terms and are excluded from that change. Check your privacy setting before you make this a routine check.

Where should a UK business report a phishing email?

Forward it to the National Cyber Security Centre's Suspicious Email Reporting Service at report@phishing.gov.uk. The NCSC says you can forward as many suspicious emails as you like, and it works to take down the scam sites behind them. Don't click any links in the message before forwarding. You don't need to forward messages your provider has already put in spam or junk. Suspicious texts can be forwarded free to 7726. If you've lost money or handed over bank details, contact your bank straight away and report it to the police's national fraud reporting service, or to Police Scotland on 101 if you're in Scotland. Tell your staff too, because phishing campaigns often hit several people in the same business.

What is the single most useful habit to stop phishing losses?

Never act on a payment or login request from the email alone. Call the sender on a number you already have, such as one from a previous invoice, your contacts or their official website typed in by hand. This one habit defeats most invoice fraud and fake supplier messages, because the attacker controls the email but not the real person's phone. Write it down as a short rule for everyone who handles money or accounts, and make it clear that nobody will be criticised for pausing to check, even if the email claims to be from the owner or a senior client.

Next stepGet this run for youWe run Claude AI, websites, booking and SEO for UK small businesses. From £495 a month.
Start a Project