AI Text Classifier was discontinued (shut down 20 Jul 2023) · last verified 13 Oct 2026
Discontinued
AI Text Classifier was shut down on July 20, 2023
AI Text Classifier
OpenAI's AI Text Classifier: the retired AI-text detector that correctly flagged just 26% of AI-written text.
Treat this page as the most useful post-mortem on AI text detection anyone has published: real numbers, real failure modes, and a shutdown instead of a sales pitch. We'd send researchers, educators, and policy teams here to understand why detectors misfire — not to detect anything. Anyone shopping for a working AI-text detector should skip it entirely and budget for provenance tooling.
Verdict generatedVerified 10/13/2026 · liveness 0/100 · cite: rightaichoice.com/tools/ai-text-classifier
- Researchers documenting the accuracy limits of AI-text detection with primary-source numbers
- Educators who want OpenAI's own caveats before trusting any detector in the classroom
- Journalists covering AI detection, false positives, and academic-integrity disputes
- Policy teams comparing detection-based screening against provenance-based approaches
- Anyone who needs a working AI-text detector today — it has been offline since July 20, 2023
- High-stakes academic or hiring decisions that need defensible evidence of authorship
- Short text, non-English text, or code classification
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip the AI Text Classifier if you need to screen text for AI authorship today — OpenAI retired it on July 20, 2023, so there is nothing to run.
There is no subscription cost, but the real cost is a false accusation: it labeled human-written text as AI-written 9% of the time.
The classifier was free throughout its short life, so there is no tier to compare against. If budget is the deciding factor for AI-text detection today, compare live detectors from established vendors rather than a retired free tool; free detectors aimed at individual teachers typically sit well below per-seat pricing from enterprise academic-integrity suites, which bundle LMS integration and reporting.
In short
AI Text Classifier — OpenAI's AI Text Classifier: the retired AI-text detector that correctly flagged just 26% of AI-written text. Best for Researchers documenting the accuracy limits of AI-text detection with primary-source numbers, Educators who want OpenAI's own caveats before trusting any detector in the classroom, Journalists covering AI detection, false positives, and academic-integrity disputes. Free to use.
What is AI Text Classifier?
AI Text Classifier — OpenAI's AI Text Classifier: the retired AI-text detector that correctly flagged just 26% of AI-written text. Best for Researchers documenting the accuracy limits of AI-text detection with primary-source numbers, Educators who want OpenAI's own caveats before trusting any detector in the classroom, Journalists covering AI detection, false positives, and academic-integrity disputes. Free to use.
OpenAI's AI Text Classifier was a free web tool for flagging AI-generated text, released January 31, 2023 and discontinued July 20, 2023 because of its low rate of accuracy. The model was a language model fine-tuned on paired human-written and AI-written responses to the same prompts, gathered from pretraining data and human demonstrations submitted to InstructGPT, with the AI side generated by several model families inside and outside OpenAI. The published numbers are the reason anyone still reads this page. On OpenAI's English challenge set the classifier caught 26% of AI-written text as "likely AI-written" while mislabeling human text 9% of the time. In the web app OpenAI raised the confidence threshold so it marked text only when very confident, trading recall for a lower false-positive rate, and told users outright not to treat it as a primary decision-making tool. Documented limits were unusually candid: unreliability under 1,000 characters, significantly worse performance on non-English text, unreliability on code, confident wrong predictions on inputs far from the training data, failure on highly predictable text, and evadability by editing AI output. OpenAI's own framing was that classifiers can be retrained against attacks but it is unclear whether detection wins long-term. For an educators and policy audience, the remaining artifacts are the methodology, the calibration warning, and OpenAI's stated pivot to provenance techniques for text plus mechanisms that show whether audio or visual content is AI-generated. If you need to screen text for AI authorship today, this tool is offline — look at provenance and watermarking-adjacent approaches instead.
Key features
- Flagged AI-written text via a language model fine-tuned on paired human and AI text (historical)
- Trained on pairs of human-written and AI-written responses to the same prompts
- Evaluated on an English challenge set at 26% true-positive rate on AI-written text
- Documented 9% false-positive rate on human-written text
- Very unreliable on inputs below 1,000 characters
- Recommended for English only; performs significantly worse in other languages
- Unreliable on code
- Could not classify highly predictable text such as a list of the first 1,000 prime numbers
- Confidence threshold raised in the web app to keep the false-positive rate low
- Prone to extremely confident wrong predictions on inputs far from its training data
- Evadable by editing AI-written text
- Free public preview launched January 31, 2023 to collect user feedback
- Discontinued as of July 20, 2023 due to its low rate of accuracy
- Publishes accuracy numbers and calibration caveats as documentation for educators and journalists
- Feedback form for teachers, administrators, parents, students, and education providers
Behind the verdict
Most retired products leave nothing behind worth reading. This one does, because OpenAI shipped the accuracy figures rather than burying them. A 26% true-positive rate on English AI-written text, next to a 9% false-positive rate on human text, tells you exactly why detection-based academic-integrity screening collapsed under scrutiny. Pick this page when you need primary-source evidence. If you're writing a policy memo on AI detection, training staff on why detector output isn't proof, or teaching calibration and base rates, the original write-up plus OpenAI's educator resource is the cleanest citation you'll find. Pass on it if you have text in front of you and need a verdict. The classifier has been unavailable since July 20, 2023, with no successor detector announced — OpenAI has said it is researching more effective provenance techniques for text and committed to mechanisms that indicate whether audio or visual content is AI-generated. Note that this is separate from the ChatGPT product line, which has since moved through several model generations; nothing in those releases revives this tool. What makes the caveats worth repeating: unreliability below 1,000 characters, degraded performance on non-English text, unreliability on code, confident errors on out-of-distribution inputs, failure on highly predictable text, and easy evasion by lightly editing AI output. That list is a checklist you can hold against any current detector's marketing claims. Where it bites: people land here from search expecting a detector and find a shutdown notice. We've seen forum threads and casual write-ups still recommend it. They're wrong, and the page itself says so. The closest thing to an alternative isn't another classifier — it's provenance. Signed generation metadata and
Limitations
- The AI classifier was retired on July 20, 2023 due to its low rate of accuracy.
- In OpenAI's own evaluations on an English challenge set, it identified only 26% of AI-written text as 'likely AI-written' while incorrectly labeling human-written text as AI-written 9% of the time.
- It should not be used as a primary decision-making tool, but as a complement to other methods of determining the source of a piece of text.
- The page notes the classifier is 'not fully reliable' and its reliability typically improves with longer input text.
as of 2026-10-03
Researching AI Text Classifier? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is AI Text Classifier still active?
AI Text Classifier was discontinued (shut down 20 Jul 2023) · last verified 13 Oct 2026.
Viability score
How well maintained and how widely used is AI Text Classifier? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: November 2026
How we score →What's new in AI Text Classifier
Checked Oct 13Across the latest 6 updates: 2 launches and 4 news mentions.
A practical guide to building with GPT-6
OpenAI published a practical guide to building with GPT-6, aimed at developers integrating the model into products.
Disrupting a coordinated model-distillation campaign
OpenAI reports disrupting a coordinated campaign to distill its models, a security item relevant to API customers.
DevDay 2026 Recap
OpenAI posted a recap of DevDay 2026, covering the developer announcements made at the event.
Introducing dots
OpenAI introduced dots, a new product listed under its product navigation.
Addendum: GPT-6.1 Sol safety
OpenAI published a safety addendum covering GPT-6.1 Sol alongside the model's release.
Introducing GPT-6.1 Sol
GPT-6.1 Sol launched, listed alongside the main model line in OpenAI's research index.
What do users say about AI Text Classifier?
What people actually say about AI Text Classifier — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
ResearchedDiscussed by users across 4 sources (YouTube, Product Hunt, Stack Overflow, Lemmy) · researched Oct 23, 2026.
Weighted by the 49 posts each of 4 sources contributed.
- +Free and accessible web tool that required no setup or technical skill
- +OpenAI published transparent accuracy numbers (26% detection, 9% false positives)
- +Candid documentation of limitations helped educators understand detection's limits
- +Raised confidence threshold in the web app to reduce false positives
- +Genuine enthusiasm from teachers and AI literacy advocates at launch
- −Caught only 26% of AI-written text, missing three out of four cases
- −Flagged human-written text as AI 9% of the time, causing false accusations
- −Discontinued after just six months, leaving users without a working tool
- −Unreliable on inputs shorter than 1,000 characters, limiting practical use
- −Performed significantly worse on non-English text, restricting global utility
- • No direct costs, but the tool's inaccuracy can lead to false accusations or missed AI use, which carries real reputational and academic costs
How we verified this
Every listed fact about AI Text Classifier — pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 13 Oct 2026. Analysis is drafted with AI models and checked against our editorial standards before publishing; see how we verify.
Verification history
We have re-verified AI Text Classifier 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed SSL certificate
- — changed pricing page, limitations
- — changed summary, pricing tiers, who should skip it, our verdict, features, who it suits, description
- — changed pricing page, limitations, models
- — changed hidden costs, our verdict, switching out, switching in, summary, description, who should skip it, setup time, pricing fit, limitations, who it suits, features, use cases, pricing tiers
- — changed pricing page, description, who it suits, our verdict, features, who should skip it, pricing tiers, summary
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does AI Text Classifier cost in 2026?
AI Text Classifier: Free to use. Pricing last verified 13 Oct 2026.
Plans compared
Every published AI Text Classifier tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Free (historical, discontinued) | Free | — | — |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where AI Text Classifier's pricing actually pencils out — and where peers do it cheaper.
The classifier was free throughout its short life, so there is no tier to compare against. If budget is the deciding factor for AI-text detection today, compare live detectors from established vendors rather than a retired free tool; free detectors aimed at individual teachers typically sit well below per-seat pricing from enterprise academic-integrity suites, which bundle LMS integration and reporting.
Has AI Text Classifier's pricing changed?
No plan price for AI Text Classifier has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick AI Text Classifier?
Pick AI Text Classifier if you are Researchers documenting the accuracy limits of AI-text detection with primary-source numbers, Educators who want OpenAI's own caveats before trusting any detector in the classroom, Journalists covering AI detection, false positives, and academic-integrity disputes, Policy teams comparing detection-based screening against provenance-based approaches.
Look elsewhere if you are Anyone who needs a working AI-text detector today — it has been offline since July 20, 2023, High-stakes academic or hiring decisions that need defensible evidence of authorship, Short text, non-English text, or code classification, Catching AI text that was lightly edited to evade detection.
Skip it if Skip the AI Text Classifier if you need to screen text for AI authorship today — OpenAI retired it on July 20, 2023, so there is nothing to run.
GeneratedReal-world workflow fit
Concrete scenarios for the personas AI Text Classifier actually fits — and what changes day-one when you adopt it.
You want a documented example of a detector's real error rates before you commit your school to any tool. You read OpenAI's published challenge-set numbers — 26% true positives on AI-written text, 9% false positives on human-written text — and the note that the web app's confidence threshold was raised to keep false positives low.
Outcome: You draft a policy that forbids using detector output as sole evidence, because OpenAI's own figures show roughly one false accusation per three catches.
A vendor claims near-perfect detection. You need a well-documented counterexample with real numbers and a named research team. You cite the January 31, 2023 launch, the July 20, 2023 retirement for low accuracy, and the paired human/AI training design.
Outcome: Your piece rests on published figures rather than vendor marketing, and you note OpenAI's shift toward provenance techniques for text, audio, and visual content.
You want a baseline. You study how the classifier was built — a language model fine-tuned on paired human-written and AI-written text on the same topic, sourced from pretraining data and InstructGPT human demonstrations — and where it broke: below 1,000 characters, non-English text, code, and inputs far from training data.
Outcome: You set your own accuracy bar well above 26% true positives and plan provenance-based signals instead of style classification alone.
Use cases
- Studying a documented example of bias and calibration failure in AI detection
- Teaching students and staff why AI detectors produce false accusations
- Sourcing real true-positive and false-positive numbers for a detection evaluation
- Reporting on academic integrity policy and the limits of text classifiers
- Comparing detection-based approaches against provenance-based methods
Setup time & first value
How long it actually takes to get something useful out of AI Text Classifier — broken out by persona, not the marketing-page minute.
There is no setup — the classifier is offline and cannot be run. For researchers reading the methodology, budgeting 20–30 minutes with the January 31, 2023 write-up covers the training design, the 26%/9% evaluation numbers, and the documented limitations. Educators writing policy should budget closer to an hour to also work through OpenAI's companion ChatGPT-for-educators resource.
Switching to or from AI Text Classifier
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From manual review: use this page's documented error rates (26% true positives, 9% false positives) as the baseline any replacement detector must beat.
- →From an older OpenAI classifier: OpenAI noted this model was significantly more reliable on text from newer AI systems than its previously released classifier — a useful reference point when evaluating current vendors.
- →From a vendor detector: compare that vendor's published false-positive rate against OpenAI's 9%, since false positives are the failure mode that hurts real people.
- ↗To a live detector: pick a tool that publishes its own evaluation numbers, because OpenAI's 26% true-positive rate shows style-based detection is difficult.
- ↗To provenance-based approaches: OpenAI says it is researching more effective provenance techniques for text and has committed to mechanisms for AI-generated audio and visual content.
- ↗To a human-review-first process: treat any detector as a complement to other methods, exactly as OpenAI instructed users to do.
Resources & guides
Tutorials & Learning
YouTube returned 6 videos for “AI Text Classifier”, and we withheld 6: 6 did not mention AI Text Classifier. We are showing none, because we could not prove any of them are about AI Text Classifier.
Official links
What are the best AI Text Classifier alternatives?
The closest alternatives to AI Text Classifier we list are Unfox AI, Humanize AI Pro and text-humanizer. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Unfox AI | Freemium | Pricing not published | — |
| Humanize AI Pro | Freemium | Pricing not published | — |
| text-humanizer | Free | Free | — |
Alternatives to AI Text Classifier
View all
Unfox AI
Unfox AI content detector flags AI-written text from ChatGPT, Claude, Gemini and DeepSeek with sentence-level analysis in 17 languages.
Humanize AI Pro
Browser-based AI text humanizer that rewrites ChatGPT output to read as human-written and clear detector checks.
text-humanizer
Free MIT-licensed Python CLI that rewrites AI text by hopping it through DeepSeek, Google Translate and optional DeepL to dodge detectors.
Tools that pair well with AI Text Classifier
Common stack mates teams adopt alongside AI Text Classifier, with the specific reason each pairing earns its keep.
Unfox AI
Unfox AI content detector flags AI-written text from ChatGPT, Claude, Gemini and DeepSeek with sentence-level analysis in 17 languages.
Humanize AI Pro
Browser-based AI text humanizer that rewrites ChatGPT output to read as human-written and clear detector checks.
text-humanizer
Free MIT-licensed Python CLI that rewrites AI text by hopping it through DeepSeek, Google Translate and optional DeepL to dodge detectors.
AI Text Classifier vs its closest rival
AI Text Classifier's closest head-to-head is with Praktika.
Our AI Text Classifier vs Praktika verdict written Jul 3, 2026 is more than 90 days old, so it is being re-reviewed against current data and is not shown.
Frequently asked questions
Categories
Used AI Text Classifier? Help shape our editorial sentiment research.