1. Red
  2. Learn
  3. WordPress

Fake captcha on my WordPress site: how to remove it

A fake captcha or 'Verify you are human' box on your WordPress site? Find where it loads from, remove it, rotate access and ask Google for a review. Dated steps.

By Timothy Lindblom, FounderPublished 4 November 20268 min readHow we check facts
On this page
  1. Why is deleting the popup not enough?
  2. Where does the injected code come from?
  3. How do I check my site, step by step?
  4. Am I affected? Where you stand right now
  5. Do I have to tell my visitors?
  6. How do I get Google's warning removed?
  7. What can I not tell from the outside?

If a fake captcha or "Verify you are human" box appears on your WordPress site, someone has added code that your pages load. Find the script and the way it got in, remove both, reset every credential that touched the site, then ask Google for a review. Deleting only the visible box fails, because the loader is usually restored.

FactDetailSource
What it isA "ClickFix" lure: a fake check that tells the visitor to paste a command into their own computerCISA AA25-203A, 22 July 2025
Scale in one campaign5,400+ compromised sites, mostly WordPress and PrestaShop, reported 5 September 2026BleepingComputer
Where the next stage livesIn a smart contract on the BNB Smart Chain testnet, not in your filesBleepingComputer, citing Netskope
How sites were enteredNot identified by Netskope at that reportBleepingComputer
Variant with TerminalVisitors sent to Windows Terminal or PowerShell, fake Cloudflare check, 28 August 2026Microsoft
Checked24 October 2026this page

This page covers the site owner's side. The visitor's side is one rule: no real verification step asks anyone to paste a command. If you also want the general sequence for any compromise, follow the step-by-step compromise response. The other WordPress guides cover related cleanups.

Why is deleting the popup not enough?

The box on screen is the last link of a chain. Netskope's campaign, as reported by BleepingComputer, injects a small script into the site. That script fetches its next stage from a smart contract on the BNB Smart Chain testnet. The contract can be changed at any time, so the text and the destination in the lure change without anyone touching your site again.

Two consequences follow. A search through your files for the words on the box finds nothing, because the words are not there. And blocking one address does not hold, since the contract can point somewhere else tomorrow.

What stays fixed is the thing on your side: a script reference or loader that you did not write, and a way for someone to place it. That is what you remove. Netskope publishes an indicator list for the campaign it tracked. Your site may be a different one, because several kits use the same lure.

Five steps from a reported fake captcha to a reviewed site: preserve evidence, close the way in, remove the script, rotate access, ask Google for review. 1 Preserve snapshot, note times 2 Close the door admins, plugins, mu-plugins 3 Remove script and loader 4 Rotate passwords, keys, salts 5 Review Google, host Order matters. Step 3 before step 2 lets the loader come back.
The removal order. Closing the way in comes before removing the script.

Where does the injected code come from?

You are looking for one of four situations. They need different fixes, and a plugin scan does not always tell them apart.

Where it loads fromWhat gives it awayWhat fixes it
Your own files (theme, a modified core or plugin file)A checksum comparison fails, or header.php, footer.php, functions.php, index.php or .htaccess changedReplace from a trusted copy by SFTP, per the WordPress.org hacked-site FAQ
A plugin you did not install, or a must-use pluginA plugin or file with an unfamiliar name and an installation date matching the first reportDelete the plugin folder or the file in wp-content/mu-plugins
A third-party script your pages embed (chat, forms, tag manager)Your files are intact, but the page's own network requests include a vendor script that changedRemove the embed, then ask the vendor for a dated incident statement
Content stored in the databaseA page or option contains script markup you did not enterSearch and remove the entry, after the way in is closed

The database row is a general WordPress possibility. No primary source we read says where Netskope's injections sit, so the rows above it lead with checks you can run.

The third row has a documented case. In September 2026, Sansec reported that scripts served from Brevo's CDN carried the overlay between 16:05:18 and 20:12:53 UTC on 14 September. Visitors saw the "prove you are human" prompt. Logged-in WordPress administrators had a plugin installed without being asked, and Sansec says it could hide from the admin list. Sansec's suggested check is to search access logs for plugin upload requests on 14 September and compare the plugin folders on disk with the admin list.

How do I check my site, step by step?

Work on a copy of your logs and a snapshot of the site, and write down times with a timezone. The WordPress.org FAQ puts documentation and a backup first, before any cleanup.

What can I do from the browser?

  • Open the site in a private window, logged out, on a different network from your office. Note whether the box shows, on which pages, and at what time.
  • Ask the person who reported it for a screenshot, the page address and the time.
  • Look at the Plugins screen, the Users screen (administrators first) and the list of Must-Use plugins. Delete nothing yet. Write down anything unfamiliar.
  • Check the Security issues report in Google Search Console, if the site is registered.

What can I do from the server?

If you have SSH or a host terminal, WP-CLI can compare files with the originals. wp core verify-checksums checks WordPress core against WordPress.org checksums. wp plugin verify-checksums --all does the same for plugins. The plugin check covers plugins hosted on WordPress.org only, so commercial and custom plugins still need a manual look.

wp core verify-checksums
wp plugin verify-checksums --all
wp user list --role=administrator
ls -la wp-content/mu-plugins

A mismatch gives you a lead to follow. Some hosts and caching plugins add files legitimately. Must-use plugins load automatically, cannot be deactivated from the admin screen, and are removed by deleting the file. Hosts use that folder too, so ask your host before deleting an unfamiliar file there.

What is the order that stops it coming back?

  1. Snapshot first. Take one more copy before changing anything, even though it is infected. The WordPress FAQ recommends it for later comparison.
  2. Close the way in. Remove administrators you do not recognise and plugins you did not install. Reset passwords for every administrator, and replace the secret keys in wp-config.php so logged-in sessions end.
  3. Remove the script. Replace changed core and theme files from a trusted copy. For core, the FAQ says to reinstall wp-admin and wp-includes by SFTP, because the dashboard installer only overwrites files and attackers often add new ones. Check the files it names: index.php, header.php, footer.php, functions.php and .htaccess.
  4. Rotate access. Change SFTP, hosting panel, database and WordPress passwords again after the site is repaired, as the FAQ advises. The full credential rotation checklist lists the rest, including API keys.
  5. Turn off the dashboard file editor. Set DISALLOW_FILE_EDIT as the WordPress hardening guide describes.

One more place deserves a look: your own computer. The FAQ points out that an infection often starts on the owner's machine, where malware can capture FTP and wp-admin logins. A full antivirus scan of your laptop belongs on the list. If the box returns after cleanup, why WordPress malware keeps coming back covers the usual second footholds.

Hello, a visitor reported a fake "verify you are human" box on our WordPress site at [time, timezone]. Please tell me: which files changed in the last 30 days for this account, whether any must-use plugin was added by you, and whether you can restore a backup from before [date]. I need the access log lines for plugin uploads and admin logins in that period.

Am I affected? Where you stand right now

Am I affected? Where you stand right now

A check from outside can see what a visitor's browser receives. It cannot see your files, your database, your administrator list or who logged in.

  1. Detected A visitor, a screenshot, a browser warning or an outside check shows a verification box asking for a command. Treat the site as compromised. Preserve a copy, close the way in, then remove the script, in that order.
  2. Unknown You see nothing, but a report exists, or the site behaves differently for some visitors. This is the common case, because lures can be shown to selected visitors. Test logged out on another network, and check the administrator list and mu-plugins folder.
  3. Confirmed by owner evidence You found the loader and the way in, and hold the file comparison and the log lines. Your site is repaired for that cause. Keep watching, because a second foothold is possible.

No scan, ours included, moves you from "unknown" to "confirmed".

Do I have to tell my visitors?

The lure works on the visitor's own computer, so the harm is theirs. In the Microsoft analysis, the original page is shown briefly and then replaced by the human-verification lure on compromised sites. Anyone who followed the instruction may have run a command on their machine.

Say so in plain words: what was shown, since when you can tell, and that no real verification step asks for a command. Suggest a full antivirus scan and new passwords for accounts used on that computer. Do not state that anyone was infected unless you know. You will not know from your side.

Under the CISA advisory, AA25-203A, the technique is a fake CAPTCHA that prompts the visitor to open the Run window and paste a clipboard command. Microsoft's TerminalFix write-up shows the same lure sending visitors to Terminal or PowerShell. So the look and the shell vary, and the rule does not: any check that asks for a pasted command is an attack.

How do I get Google's warning removed?

If Search Console shows a hacked-content or malware warning, Google's Security issues report says to fix the problem on all affected pages and then select Request Review. A partial fix earns no partial return. Google says most reviews take several days or weeks.

Write the request as a short account: what you found, what you changed, and the date. Wait for a decision before sending another.

What can I not tell from the outside?

Plenty. A free scan reads the technologies your site serves and matches versions to public vulnerability records. It is a lookup, and it does not read your database or your administrator list. The page on where outside scans stop short explains the limits.

What an outside view can still add: whether .git, .env and backups are readable, which versions of WordPress and plugins are served, and whether panels or debug pages are left online. Netskope could not say how its 5,400 sites were entered. An out-of-date plugin is a candidate, since Microsoft describes drive-by compromises that are often reached through WordPress vulnerabilities. An unfamiliar administrator, as covered in hidden admins from a form plugin flaw, is another. Both are for you or your host to confirm.

If you want a full outside look at what a site exposes, the full scan is $100 per scan, and the free scan is $0, one per domain.

Questions people ask

Is a fake captcha on my WordPress site a sign that I was hacked?

Usually yes. A real captcha never asks a visitor to paste a command, so a box that does was added by someone who can change what your pages load. It can come from your own files, your database, or a third-party script you embed. Treat it as an incident until you know which.

I deleted the popup code and it came back. Why?

Removing what you can see leaves the way in open. The injection is usually put back by a second foothold, such as an extra administrator, a plugin you did not install, a file in the must-use folder, or a stolen password. Find and remove the foothold first, then the script, then rotate access.

Can I see the box myself? Visitors say it appears but I never see it.

Often not. Microsoft describes the real page being shown briefly before the lure replaces it, and campaigns can show it to some visitors only. Test in a private window while logged out, from a different network and device, and ask the visitor for a screenshot and the time.

Do I need to tell visitors who saw it?

If a visitor followed the instruction and ran a command, their computer is the problem, not your site. Tell them plainly what was shown and since when, and suggest they run a full antivirus scan and change the passwords they use. Keep the wording factual and avoid saying anyone was infected.

How long until Google stops warning about my site?

Google says most reviews take several days or weeks, and you request one only after every page is fixed. Fixing some pages does not earn a partial return. Resubmitting before a decision does not speed it up.

Will a security plugin scan find it?

It may, and it may not. A scanner running inside a compromised install is only as reliable as the files it reads. Use it as one signal next to a checksum comparison, a look at the administrator list and the must-use folder, and the rendered page from outside.

Sources

  1. BleepingComputer: Over 5,400 hacked sites serve ClickFix payloads stored on the blockchain (5 September 2026)Report on the Netskope findings: scale, WordPress and PrestaShop, blockchain-hosted next stage, entry point unknown. Checked 24 October 2026.
  2. Netskope Threat Labs: Malware on the blockchain, an ongoing campaign's new WebRTC twistThe original research. The page body was truncated in our fetch, so figures are taken from the BleepingComputer report above.
  3. Netskope Threat Labs IOC repository (Amatera ClickFix)Indicator list published by Netskope for this campaign.
  4. Sansec: Brevo supply chain attack (16 September 2026)A third-party script on customer sites showed the overlay and installed a plugin for logged-in admins. Checked 24 October 2026.
  5. Microsoft Security Blog: Think before you Click(Fix) (21 August 2025)How the verification lure works and that it is delivered from compromised websites.
  6. Microsoft Security Blog: TerminalFix campaign (28 August 2026)A variant that sends visitors to Terminal or PowerShell, delivered from compromised sites with a fake Cloudflare check.
  7. CISA, FBI, HHS and MS-ISAC: #StopRansomware Interlock, AA25-203A (22 July 2025)Government description of the fake CAPTCHA technique.
  8. WordPress.org: FAQ My site was hacked (last updated 26 July 2026)Document, snapshot, reset access, replace core folders by SFTP, files to inspect. Checked 24 October 2026.
  9. WordPress developer docs: Hardening WordPress (modified 14 October 2026)DISALLOW_FILE_EDIT and file permissions.
  10. WP-CLI: wp core verify-checksumsCompares core files with WordPress.org checksums.
  11. WP-CLI: wp plugin verify-checksumsCompares plugin files with WordPress.org checksums. Works for plugins hosted on WordPress.org only.
  12. WordPress developer docs: Must-Use Plugins (modified 14 October 2026)Loaded automatically, not switched off from the admin screen, removed by deleting the file.
  13. Google Search Console help: Security issues reportFix every page, then Request Review; reviews can take several days or weeks.

Facts on this page are dated and point at the primary source next to them. If a vendor status changes after the date shown, the vendor's page wins; tell us at support@run.cloud and we will update it.