Copyleaks on October 14, 2026 released Copyleaks Content Guard, a WordPress integration that runs AI detection and plagiarism checks on articles inside the content management system before they are published, and pairs it with a service that searches the web for unauthorized reuse of a publisher's journalism.

In Short

Copyleaks made a WordPress add-on that checks whether an article looks like it was written by AI, or copied, before an editor hits publish. That matters to newsrooms and content teams that have written rules about AI use but no way of checking every story against them, other than asking writers to own up. What changes is that the check happens automatically inside the editing tool, although the pricing and the accuracy figures come only from Copyleaks and the company's own web pages do not fully agree with each other.

What Copyleaks released

The New York-based company described Content Guard as a solution "that helps publishers protect their journalism throughout its lifecycle." According to Copyleaks, the product is "already available for WordPress" and brings the company's AI Detection and Plagiarism Detection technology "directly into editorial workflows." WordPress is the first CMS supported. Integrations with "other leading content management systems and publishing platforms" are planned, according to the company, which gave no names and no dates for them. Publishers on other systems can build the same checks through the Copyleaks headless API, "customized to fit their existing editorial workflows."

There are two halves to the product, and Copyleaks frames them as upstream and downstream. The upstream half sits inside WordPress and evaluates drafts "before it goes live, in accordance with each publisher's own AI editorial policies." The downstream half is a web scan. According to Copyleaks, it searches "60 trillion web pages for plagiarism and unauthorized reuse" across multiple languages, so that publishers can "identify plagiarism, protect intellectual property, uncover licensing opportunities, and safeguard the commercial value of subscription-driven content."

Alon Yamin, CEO and co-founder of Copyleaks, put the pitch in commercial terms. "AI is reshaping how content is created, but trust remains the foundation of publishing," Yamin said. "The publishers that succeed in the AI era will protect both the integrity and the commercial value of their content. Content Guard helps organizations make informed decisions about the quality, originality, authenticity, and human authorship of every article, while giving them greater visibility into how their journalism is reused across the internet."

A naming gap between the release and the product page

The release points readers to copyleaks.com/copyleaks-wordpress-plugin for more information. That page, as captured, does not use the name Content Guard anywhere. It calls the product the "Copyleaks WordPress Integration" and the "Copyleaks WordPress Plugin," and pitches it as "Enterprise AI and Plagiarism Detection built directly into WordPress." The page also lists a grammar checker as a third component, which the release does not mention, and it says nothing about the 60-trillion-page web monitoring service. Whether the web scan is part of the plugin or a separate Copyleaks service is not stated in either document.

How the WordPress plugin works

The technical detail sits mostly on the product page rather than in the release. According to Copyleaks, the plugin has four core features.

Native, real-time analysis shows "live AI, plagiarism, and grammar metrics right inside the standard WordPress post grid." A screenshot Copyleaks circulated with the launch shows what that looks like in practice: an extra column labelled "Copyleaks AI Score" in the WordPress Pages list, next to the title, author and date columns. Three sample pages in the screenshot show a score of 0%, and a fourth shows the status "Scan failed." Copyleaks did not explain what triggers that status.

"No-Lift" automation is the feature that makes the system work without editors remembering to run it. According to the product page, the integration "automatically triggers scanning the second a post is put into 'Pending' status." Pending is the WordPress state a contributor uses to submit a draft for review, which is the point at which an editor would normally pick a piece up. Tying the scan to that status means a score exists before a human editor opens the story.

A centralized scan center gives admins a dedicated dashboard tab from which they "manage site-wide settings, configure security thresholds, and run historical deep-dives across existing pages," according to Copyleaks. The historical scan is notable: it implies a publisher can run the detector over its back catalogue, not just new submissions.

Flexible role permissions let administrators "control exactly which users can view individual scores, adjust scanner sensitivity, or bypass content rules." A newsroom could, in principle, let editors see scores while hiding them from freelance contributors, or let senior staff skip checks altogether. The page does not say how the bypass is logged.

On setup, Copyleaks claims a "rapid 15-minute setup": upload the plugin, enter API credentials in the settings tab, and start scanning, with "no complex development or engineering resources required." That API credential step matters for cost, which is discussed below.

What the detector says it does

Content Guard is powered by what Copyleaks calls AI Logic, a feature the company says "helps explain why content is flagged as AI-generated." According to the release, AI Logic is "trained on trillions of human-written documents" and combines several methods, including AI Phrase analysis and AI Source Match, "to identify statistical patterns left behind by leading LLMs." The product page puts the training figure slightly differently, as "trillions of pages of human text." Neither document gives a precise number.

The release presents AI Logic as a way to get "multiple analytical signals, not just a single AI score." That is a response to a familiar complaint about AI detectors: a bare percentage gives an editor nothing to argue with or check. Copyleaks' launch outreach said editors see "how much of an article appears AI-generated and the specific copy flagged," which suggests highlighted passages rather than a verdict alone.

The product page adds one sentence on method that is worth reading closely: "AI detectors can be trained on human-written content samples, so the presence of an AI text generator becomes clearer compared to human writing." In other words, the model learns what human writing looks like and flags deviation from it. Copyleaks did not publish false positive rates, test sets, or how it treats text that has been AI-assisted and then edited by a person.

The vendor's numbers

Several figures in the release and on the product page are vendor-supplied and were not independently verified for this article:

  • AI detection "with over 99% accuracy, across the leading LLMs, including ChatGPT, Gemini, DeepSeek, and Claude," according to the release. No methodology, sample size or false positive rate accompanies the figure.
  • A web index of 60 trillion pages for plagiarism and reuse scans, according to the release.
  • "Trusted by 450+ global enterprises," according to the product page, which also says the plugin can scan "thousands of text files in a single workstream."
  • A CODiE Award for Best AI-Driven Technology and a place on Inc.'s Best in Business list, according to the product page, and a No. 424 ranking on the 2026 Inc. 5000 plus a listing on Fast Company's World Changing Ideas, according to the release.

Languages: three different numbers

The language claims do not line up across the three documents. The release says Content Guard can "detect plagiarism and analyze content across more than 30 languages." The product page says the plugin "enables robust cross language plagiarism detection in 30+ languages." The Copyleaks pricing page, however, lists "AI detection in 30+ languages" and "Plagiarism detection in 100+ languages" for its Personal plan. The most likely reading is that AI detection covers 30-plus languages and plagiarism detection covers more, but the WordPress materials state the lower figure for plagiarism, and Copyleaks has not clarified which applies to the plugin.

What it costs, and what is not published

Neither the release nor the plugin page gives a price for Content Guard. The plugin page directs visitors to "Get Started," "Get the Plugin" and "Get a Demo," and its language - "enterprise-grade," "high-volume content operations, digital media newsrooms, agency networks, and media organizations" - places it in Copyleaks' Enterprise tier. The pricing page has a separate Enterprise tab, but the captured version shows only the Individual plans, with no Enterprise price.

The Individual plans give some sense of how Copyleaks meters usage. On annual billing, which the page says saves "up to 25%," the Personal plan costs $13.99 a month, billed as $167.88 a year, and includes 1,200 unified credits, enough to "scan up to 300,000 words OR 1,200 images." The Pro plan costs $74.99 a month, billed as $899.88 a year, and includes 25 user seats, "designed for startups and small teams," with 12,000 unified credits for up to 3,000,000 words or 12,000 images. Both ratios work out to 250 words per credit. Pro also adds advanced detection filters, scanning of "entire websites," cross-language translation detection and an analytics dashboard. Extra credits can be bought at any time, according to the page, and "taxes may apply."

None of the Individual plans lists the WordPress plugin among its features. The Personal plan names a browser extension and a Google Docs add-on; Pro inherits those. Because the plugin runs on API credentials, a publisher's cost will depend on its API allocation and on how many words it pushes through the scanner each month, not on a flat plugin fee. A newsroom publishing 50 articles a day at 800 words each would put roughly 1.2 million words a month through the Pending-status trigger, before counting rewrites that re-enter the queue. That arithmetic is illustrative; Copyleaks has not said whether a resubmitted draft consumes fresh credits.

The pricing page also carries a banner for "AI Image Detection" as new, and the plugin page's header announces that "AI Video Detection is now live." Neither the release nor the plugin page says whether images or video embedded in WordPress posts are scanned. The plugin materials refer only to text.

Why publishers are being sold this now

The release identifies two problems. Internally, publishers "must establish and consistently enforce editorial standards governing AI use by staff writers, freelancers, contributors, and syndicated content." Externally, they "must protect original journalism from being copied, paraphrased, or repurposed as AI makes it faster and easier to produce low-quality derivative content at scale."

The internal problem gained a public example this summer. In its outreach for the launch, Copyleaks cited the controversy over a Wall Street Journal op-ed by investor Stanley Druckenmiller. The essay, "Let the Bond Market Speak," ran on August 24, 2026, and was flagged as AI-generated by a detection tool from Pangram, a Copyleaks competitor; Druckenmiller then confirmed he had used AI and said he was not embarrassed by it, according to The Wrap. WSJ editorial page editor Paul Gigot defended the piece as Druckenmiller's genuine opinion. The episode showed that outside detectors will be pointed at published work whether or not the publisher runs one itself.

The regulatory pressure is also real. The European Commission published guidelines and a Code of Practice on AI-generated content transparency in July, with Article 50 AI Act obligations applying from August 2, 2026, and fines of up to EUR 15 million or 3% of worldwide annual turnover. Google, for its part, added language to its Search Central guidance on October 1 warning that AI output can contain hallucinations and should be checked before publishing, covering title tags, meta descriptions, structured data and alt text as well as body copy. A week later, PPC Land reported that Google's people-first content guidance now treats fabricated author profiles as a sign of an untrustworthy page. Mass-produced pages fall under Google's scaled content abuse policy whether the text comes from people or machines.

For advertisers, the same question has a supply-side form. A TAG, ANA and Fiducia analysis found that AI slop inventory scores as higher quality than clean supply on standard verification metrics, estimating it at 1.3% to 2.4% of open web programmatic spend - close to the share held by made-for-advertising sites. Earlier, DoubleVerify flagged AI-generated recipe sites drawing millions of monthly visits, and IAS identified AI slop sites as a major ad quality threat. Publishers that want to sell premium inventory have a commercial reason to show their own pages are not part of that pool. Copyleaks makes that argument implicitly when it promises to help publishers "demonstrate to readers when original reporting is human-written."

The reuse side

The downstream web scan targets a different fight. Publishers have spent two years trying to get paid when AI systems use their work. Penske Media sued Google in September 2025 over the use of its content in AI features. IAB Tech Lab put out its Content Monetization Protocol for comment in March 2026, a framework requiring AI systems to have commercial agreements before crawling. In June, 31 UK sites began billing AI scrapers a typical £500 per article through county small claims courts.

All of those efforts need evidence of reuse. A crawler log shows that a bot visited; it does not show that a paraphrased version of an investigation ended up on someone else's site or in a model's answer. Copyleaks' web scan is positioned to supply that second kind of evidence, and the release explicitly mentions "licensing opportunities." Whether its output would stand up as evidence in a dispute is a separate question the company does not address.

Questions the materials leave open

Several practical details are missing from the documents Copyleaks supplied.

What happens when a draft crosses a threshold? The release talks about helping "prevent unauthorized or non-compliant AI-generated content from being published," and the product page mentions "security thresholds" and the ability to "bypass content rules." Neither says whether the plugin can block a post from moving out of Pending, or only displays a score. The screenshot shows a score column, not a block.

How are mixed workflows handled? The release says the tool can apply AI guidelines "whether AI is permitted for research and editing or subject to stricter limitations." Text that a writer drafted and then polished with an AI tool is exactly the case where detectors are weakest, and Copyleaks gives no figures for it.

Where does the text go? The plugin sends draft content to Copyleaks through its API. Unpublished journalism, including embargoed material and source-sensitive reporting, would leave the publisher's infrastructure before publication. The documents do not describe retention, whether submitted drafts feed the company's Shared Data Hub (which the Personal plan advertises "for enhanced detection accuracy"), or how a publisher can opt out of that.

How does it treat syndication? The release lists syndicated content among the material editors can standardize review for. Syndicated wire copy will, by design, match text elsewhere on the web, and a plagiarism scan run against it will produce matches. The documents do not explain how the plugin distinguishes licensed copies from copying.

And what does 99% mean? An accuracy figure for an LLM detector can be computed on balanced test sets that look nothing like a newsroom's queue. A detector that is right 99% of the time and runs on 1,500 stories a month would still misclassify around 15 of them. For an editor deciding whether to confront a freelancer, the false positive rate on human text matters more than the headline accuracy, and that rate is not published.

Timeline

Summary

Who: Copyleaks, a New York-based AI detection and plagiarism company led by CEO and co-founder Alon Yamin, aiming the product at publishers, newsrooms, agency networks and media organizations running WordPress.

What: Copyleaks Content Guard, a WordPress integration that automatically scans posts for AI-generated text, plagiarism and grammar when they enter Pending status, shows a Copyleaks AI Score in the post list, and is paired with a web scan the company says covers 60 trillion pages for unauthorized reuse. Accuracy and scale figures are vendor-supplied, no plugin price is published, and the release, product page and pricing page differ on the product name and plagiarism language coverage.

When: Released October 14, 2026, and described by Copyleaks as already available for WordPress, with other CMS integrations planned but undated.

Where: Inside the WordPress CMS, with a headless API for publishers on other platforms, and the reuse scan running across the open web in multiple languages.

Why: Publishers are writing AI policies for staff and freelancers without a systematic way to enforce them, while regulators, Google's search guidance and advertisers are all paying closer attention to AI-generated content and to the reuse of original journalism by AI systems.