Writer.com AI Detector: What Happened to the Free Tool

Does Writer.com Still Have an AI Content Detector?

No, not as a public tool at the address it lived at. Writer, Inc. is an enterprise generative-AI company, and on August 10, 2026 the URL of its free AI Content Detector, writer.com/ai-content-detector/, returned a 301 redirect pointing at an anchor on the company homepage. The disclaimer page that the tool linked to returned the same redirect in the same second. The redirect goes nowhere useful: I fetched the live homepage and it carries no element with that anchor and not a single occurrence of the string detector. Writer’s own page sitemap, 165 URLs long, lists no detector page at all (EV-07).

The dates line up with the company’s own announcements. A banner on the tool page told visitors it would be sunset on December 22 along with the AI detection API endpoint, and that it would work normally until then (EV-01). That banner named no year, which is why an uncited shutdown claim naming the following year, still circulating on one alternatives post, has been easy to repeat (EV-14). Writer’s developer documentation settles it: the deprecation notice for the detection endpoint states December 22, 2025 explicitly, and I read it in two separate archive captures taken months apart (EV-02). The capture history agrees. The tool page answered normally on December 5 and December 9, 2025, then redirected from December 24 onward, with a 404 recorded in February 2026 and redirects continuously through August 2026 (EV-07).

One honest limit belongs here rather than in a footnote, because it is the only thing on this page a Writer employee could correct me on. A JavaScript application shell still answers at a subdomain of writer.com built for a content-platform integration. I fetched it: HTTP 200, a page title, and an empty application root with a message asking you to enable JavaScript (EV-07). Whether it still returns scores to anyone who loads it in a real browser, I could not establish, because checking it needs a browser I did not have. So the defensible statement is the narrow one: the public page and its URL are gone, and no detector appears in Writer’s navigation or its sitemap. If you need a check done today rather than an answer about what happened, the detector we run at MeteGPT is open, free, and asks for no account.

What Was Writer.com’s AI Content Detector?

Writer.com’s AI Content Detector was a free web tool that took a passage or a URL, ran it through a classifier, and returned one number. You pasted text into a box or supplied an address, pressed a button labelled Analyze text, and got a percentage split between AI-generated content and human-generated content (EV-06). There was no sentence-level highlighting and no per-passage report; nothing on the page offered an export or a plagiarism check. What made it easy to reach for was the friction, or rather the absence of it: no email, no account, no card, and a 5,000-word ceiling (EV-03).

Who Is Writer, Inc.?

Writer, Inc. sells an enterprise platform, and understanding that explains more about this detector than any accuracy argument does. Its encyclopedia entry describes a generative artificial intelligence company based in San Francisco offering a full-stack platform for enterprises, and does not mention a detector anywhere (EV-09). The archived tool page sat inside that company’s navigation: a platform menu listing agents, a studio, in-house language models and a knowledge graph, a departments menu reading Finance, HR, Legal, Marketing, Sales and Support, and Request a demo as the header call to action (EV-08). The detector itself carried no price and no plan of its own. That last detail is also a correction owed to two of the reviews still coming up for this search, which present a per-seat platform subscription as though it were the detector’s paid tier. A humanizer vendor’s own review states the point plainly: the free version was the only version, and paid Writer plans did not upgrade the detector (EV-16). So a free consumer tool sat on a sales-led business suite. I am describing the structure, which is documented, and not the reasoning, which is not.

Is Writer.com the Same as WriteHuman or Writesonic?

No. Writer, Inc. and WriteHuman are separate companies doing opposite jobs: Writer, Inc. ran a detector that scored text, and WriteHuman sells a humanizer that rewrites it. Nothing I verified suggests any relationship between them in ownership, technology or naming, and the resemblance is a coincidence of English vocabulary. The same applies to Writesonic, which is a separate AI-writing company with its own detector.

The confusion is worth heading off because it is unusually easy to fall into here. Four other names crowd the same letters: StealthWriter and NeuralWriter, both unrelated companies; the domain aidetectorwriter.com, which is a different product on a similar-looking address; and a mobile app whose name also pairs detector with writer. None of them is Writer, Inc., and no rating, review or accuracy figure belonging to any of them can be attached to the tool this page is about. There is also the plain English word: a search for the bare brand token returns mostly a large corporate group with no connection to any of this. Everything below concerns Writer, Inc. at writer.com and nothing else.

Was Writer.com’s AI Detector Accurate?

Nobody can tell you, and the interesting part is why. Writer, Inc. itself never published a figure. I read the full rendered text of two archive captures of the tool page, five months apart, and scanned both for every word the claim would have to contain: no accuracy percentage, no precision or recall, no benchmark score, no dataset, no named validation, no third-party test, no model name (EV-06). The only percent sign on the page was the empty placeholder waiting for your result. A published validation study that included this detector as one of five tested tools reached the same conclusion independently, recording that no information from credible sources regarding its methods was available at the time of writing (EV-10). That study’s own headline figures are pooled across all five detectors it tested and carry no per-detector breakdown, so none of them describes Writer.com’s tool and I am not going to borrow them.

What does exist is one structured test, published by a company that sells a competing detector. Its stated design was seven short articles generated with a GPT-3-era tool, run through both products. The reported result for Writer.com is a miss pattern rather than a score: one of the seven samples flagged as substantially AI, and on two of them, in the publisher’s own words, it “failed to detect any AI content at all” (EV-11). Three limits ride with that. Seven samples is a small run. A single generator from one model generation is a narrow input. And with no human-written control set, the test can only show missed AI text and says nothing at all about the error most readers actually fear, which is human writing flagged as machine-made. The averages that test produced are the figures I would normally quote here, and I will not, for a reason that gets its own section further down.

There is a fourth number in circulation, and it deserves naming precisely because it looks the most authoritative. One review site publishes an accuracy percentage, a false-positive rate and a false-negative rate for this detector, attributed to a benchmark it dates to March 2026. That is three months after the tool’s documented removal. The page carries no author, no publication date, and, when I scanned its outbound links, no external citation of any kind (EV-18). I cannot reconcile a benchmark with a tool that was not reachable, so I am reporting the composition rather than the figures, and the figures do not appear on this page.

What Writer, Inc.’s Own Disclaimer Said About Its Detection Scores

Writer, Inc. kept a separate disclaimer page for this tool and linked it from the tool itself with the words “Read the disclaimer first.” Almost nothing written about this detector quotes it. Under a heading about accuracy it said: “Our AI-generated text detectors try to be accurate, but it’s hard to be 100% accurate.” Under a heading about predictions it said: “The prediction we give is just for showing how our product works. It’s not meant to be used as a real product.” The page’s own header dates its revision to January 2024 (EV-04). One click away, the tool page invited readers to check their work and “decide if you want to make adjustments before you publish” (EV-03). Both sentences were live on writer.com in captures three days apart in December 2025. I am recording that both existed, one click apart. I am not claiming anyone intended the gap.

The same disclaimer also explains the miss pattern that the outside test found, in the vendor’s own words rather than mine. Writer, Inc. wrote that it focused on “having a very low number of false positives,” describing that as important for “stopping false accusations in situations like academic honesty cases” (EV-05). A classifier deliberately tuned to almost never accuse will, mechanically, let more machine-written text through. That is a design trade-off rather than incompetence, and it reframes the two samples the tool returned nothing on. The same page advised against using Writer, or any detector, without a human reviewing the output, and stated that the product worked better on English than on other languages (EV-05).

Credit where it is owed: Writer, Inc. published that caveat at all, and one of the reviews still coming up for this search did surface it, reporting that the vendor described its predictions as being for demonstration purposes and unable to guarantee accuracy (EV-19).

Was the Writer.com AI Detector Free, and What Was Its Word Limit?

It was genuinely free, and it accepted up to 5,000 words per submission. The tool page said so in its own words, “Use our free AI detector to check up to 5,000 words,” and the counter beside the input box read the same figure. Two captures six months apart carry identical wording, and the full page text shows no signup field, no email capture and no pricing (EV-03). Two further reviews corroborate the 5,000-word ceiling and the absence of an account requirement: one from a humanizer vendor (EV-16), one from an affiliate-disclosing review blog that links commercial alternatives (EV-15).

That number matters because two of the three top-ranking reviews get it wrong. Two of them state a cap of 1,500 characters, roughly a fifth of what the vendor’s own page advertised, and a third states no cap at all (EV-11, EV-12, EV-13). It is a small error, and it is the cheapest available demonstration that the writing about this tool was not done with the tool open.

Since I am about to point you at my own product, here is its ceiling in the same breath. MeteGPT’s free tier gives you 125 words a run, four runs a day, with no account. Against a 5,000-word no-signup box that is not a slightly smaller version of the same thing, it is a different shape of product: enough to see how a paragraph reads and how the scoring behaves, nowhere near enough to screen a 1,200-word post or a client batch. If your checking is a scheduled part of the work rather than a one-off, that daily ceiling is exactly where the free tier stops, and where that 125-word free ceiling stops and the paid tiers begin is written out rather than buried behind a signup.

Did Writer.com’s AI Detection API Shut Down Too?

Yes, on the same date. Writer’s own banner said the tool was going “along with the AI detection API endpoint” (EV-01), and the developer documentation for that endpoint carried a deprecation notice naming December 22, 2025 and advising integrators to migrate (EV-02). The documentation URL now redirects away as well. If you wired that endpoint into a publishing or moderation workflow, it stopped answering on a known day, and none of the reviews I read mentions it. The honest answer from my side is a refusal: MeteGPT does not sell a detection API, so we are not a drop-in replacement for that integration. What we run is a check you perform yourself and a rewrite that comes back with a score attached.

Why Do the Writer.com Reviews All Report the Same Score?

Because two of the top-ranking reviews carry the same numbers. A competing detector vendor and a humanizer vendor both published seven-sample tests of Writer.com’s detector, and the decimals match across the two pages while the winning brand changes. Here is what each states, quoted from pages I fetched in the same pass on August 10, 2026.

Two things follow, and only two. The stated inputs are mutually incompatible, so both descriptions cannot be true of the same numbers, and one set of figures cannot have come out of two different generators. And the direction is undetermined: the humanizer vendor’s page carries the earlier publication date of the two, which rules out the assumption most readers would reach for first. I am not asserting that either publisher copied anything, and nothing in what I gathered establishes intent. What is publishable is the arithmetic: the one number a reader meets everywhere in this result set has provenance that cannot be established, which is why the accuracy section above cites the miss pattern from that test and not its averages. The same standard gets applied to that publisher’s own product in my dated record of Originality.ai.

What Does an Old Writer.com Detection Score Mean Now?

Less than whoever is holding it probably thinks, and this cuts in both directions. A screenshot of a Writer.com detection score is an artifact of a classifier that never published an accuracy figure, a precision or recall number, a benchmark or a dataset (EV-06, EV-10), whose own vendor described its output as a demonstration rather than a product (EV-04), and which was deliberately tuned to accuse as rarely as possible (EV-05). It also cannot be re-run. The tool is not reachable, so neither you nor the person quoting it at you can reproduce that reading or have a third party confirm it. A clean score from it was never a clearance, and a flagged one was never a finding.

If you are on the receiving end, the practical moves are procedural rather than technical: keep your draft history and version records, ask which system produced the score and on how many words, ask to see the specific passage in question, and get a second dated reading from something that is currently running. If curiosity has taken you toward rewriting your way out of a flag instead, that is a different subject with its own risks, and I have written about what actually happens when people try to get around a detector separately rather than in passing here.

What If It Flagged Writing You Did Yourself?

Then you have run into the failure mode this whole category is worst at, and Writer, Inc. named the relevant limitation itself: its product worked better on English than on other languages (EV-05). Peer-reviewed work supports treating that as a category-wide problem rather than a quirk of one vendor. A Stanford-led study measured heavy false-positive rates on TOEFL essays written by non-native English speakers across the detectors it tested, and Writer.com’s tool was not among them, so that finding is background for the class and never a figure for this product (EV-best-ai-humanizer-01). If the piece in question is an essay you wrote and a checker disagrees, the ground to stand on is documentation, and I have set out what to do when a checker flags an essay you wrote yourself in more detail.

What Should You Use Instead of the Writer.com AI Detector?

The landscape moved on without replacing what this tool was, which was a large free box with no login. On the evidence in our own dated record, current detectors mostly either meter you or want an account, and the ones worth considering differ mainly in what they disclose about themselves. Rather than rank them again here, I keep the detectors that are still running, ranked by what the evidence shows as a separate, dated record.

For our own side of it, here is the plain version. MeteGPT runs a detector you can open without an account and a humanizer that returns a rewrite with a detector score attached to it on the same screen, which is the part people tend to find useful: you see how the change landed rather than guessing. The engine behind it was trained on 2,590 real student essays, and every rewrite is scored by a trained detector before it comes back to you. Where the free tier stops is written above, and MeteGPT’s rewrite-and-score screen is the fastest way to see whether any of that is useful to you.

Two limits stated out loud, because a page criticising other people’s disclosure should manage its own. We publish no accuracy figure for our own detector, for the straightforward reason that nobody outside this company has measured it, and I would rather say that than print a number I could not defend. The measured result we do publish, FC-METEGPT-001, is a small controlled in-house run from mid-May 2026 that scores our humanizer’s output against six named detectors, with Turnitin carried only as an under-20% bound because Turnitin’s own report shows an asterisk instead of a number below that threshold (EV-turnitin-10). It measures how our rewriting engine’s output reads, not how well our detector judges anyone’s writing, and those are different claims that this page keeps apart. There is no Writer.com row in that matrix and there never can be one, because the tool is decommissioned and therefore permanently unmeasurable. The full figures, and every caveat that travels with them, sit in the record behind that measurement.

Was Writer.com a Good AI Detector? The Verdict

It was a genuinely useful free tool with an unusually honest disclaimer and no evidence base whatsoever, and it is now unreachable at its published address. Held to what it actually was, Writer.com’s AI Content Detector took 5,000 words for nothing and asked for no account. Held to what a score from it could support, it never published an accuracy figure across its entire life, an outside study looking for its methods found none, and its own disclaimer described the prediction as a demonstration rather than a product. Those two readings are not in conflict. A free thing that tells you it is a demonstration is behaving reasonably; the problem is that its output kept getting quoted as though it were a finding.

The last thing worth saying is about the writing around it rather than the tool. Every detailed review of this detector that comes up for the search, with one exception, has something to sell in the same market or is paid to point at someone who does, and the exception carries no author at all. At least four were edited after the tool’s removal date and still describe it in the present tense (EV-12, EV-13, EV-15, EV-16), and none of the ones I read says it is already gone. I am in that same category of publisher, which is why my conflict sits at the top of this page rather than the bottom, and why every claim above carries a date and a link rather than an assurance. Others noticed the retirement before I did (EV-14, EV-17), so this is not a scoop and I am not claiming it as one. The only thing I would claim is method: you can open every source here and check me. If what you need is the thing this page is about the absence of, run one paragraph through our own checker and read the score yourself, and if the checking is going to be a regular part of your week, the plan that lifts the daily cap is the honest next step.

Common Questions

Is the Writer.com AI detector still available? Not at its published address. On August 10, 2026 writer.com/ai-content-detector/ redirected to a homepage anchor that does not exist, and no detector appears in Writer’s navigation or its 165-URL sitemap (EV-07). Writer, Inc. announced the retirement on the tool page itself and dated the API removal to December 22, 2025 in its developer documentation (EV-01, EV-02).

Was the Writer.com AI detector free, and what was the limit? Free, with no account, no email and no card, up to 5,000 words per submission according to the vendor’s own page and its own counter (EV-03). The 1,500-character figure carried by two of the three top-ranking reviews does not match what Writer published (EV-11, EV-12).

How accurate was Writer.com’s AI detector? No trustworthy figure exists. Writer, Inc. published none in the tool’s entire life (EV-06), an independent study looking for its methods found nothing from credible sources (EV-10), and the one number in wide circulation appears on two vendor pages with incompatible stated inputs (EV-11, EV-12).

Is Writer.com related to WriteHuman? No. They are separate companies in opposite roles: Writer, Inc. ran a detector, WriteHuman sells a humanizer. Nothing verified here suggests any relationship of ownership, technology or naming, and the same goes for Writesonic, StealthWriter and NeuralWriter.

Did the Writer.com AI detection API shut down as well? Yes, on the same date. Writer’s banner named the endpoint alongside the tool (EV-01) and the endpoint’s own documentation carried a deprecation notice for December 22, 2025 (EV-02).

Does an old Writer.com score still mean anything? It cannot be re-run or confirmed by anyone, since the tool is not reachable. It came from a classifier with no published accuracy record that its own vendor called a demonstration rather than a product (EV-04, EV-06), tuned to accuse as rarely as possible (EV-05).

Corrections and updates. Two specific errors this page corrects, both checkable against primary sources: the shutdown year, which one page-one post places a full year later than Writer’s own documentation and cites no source for, while that documentation names December 22, 2025 (EV-02, EV-14), and the free word cap, which two top reviews give as 1,500 characters while Writer’s own page says 5,000 words (EV-03). If either correction turns out to be wrong, it gets fixed here in place with the date of the fix attached, not quietly edited away.