GPTZero added a Hallucination Detector, and it does something real: it flags convincing but non-existent citations, searching more than 220 million scholarly articles to catch sources an AI invented. If your problem is a made-up reference, it helps. But GPTZero is careful about what it is not, and that boundary is the whole story here.
GPTZero's own FAQ states the tool 'does not take a stance on whether your claims or the claims in the sources cited are true.' Its founder has called the related Sources feature 'explicitly not a fact-checker.' So it can tell you a citation exists; it cannot tell you the sentence it supports is correct. This guide maps that boundary precisely, and shows what covers the other side of it.
What GPTZero's hallucination tool actually checks
GPTZero started as, and fundamentally still is, an AI-content detector: it estimates the probability that text was written by a machine. Around that it has built citation tools. The Hallucination Detector highlights sentences likely to contain fabricated or incorrect citations and points you to real academic sources instead. The related Sources feature detects factual claims, matches them to online and academic references, marks whether the evidence supports, contradicts, or debates a claim, and even runs AI-detection on the cited sources to catch second-hand hallucinations.
That is genuinely useful for one specific failure: a citation that does not exist. What it is not built to do is judge truth. In GPTZero's own words, the tool 'does not take a stance on whether your claims or the claims in the sources cited are true.' It checks that a reference is real and roughly on-topic, not that your statistic is right, your quote is accurate, or your conclusion follows. Catching an invented citation and verifying a claim are different jobs, and GPTZero only signs up for the first.
GPTZero vs verification, side by side
Both touch citations. Only one is about whether you are right.
| Dimension | TrueStandard | GPTZero |
|---|---|---|
| Core question | Is what this says true? | Was this AI-written, and do its citations exist? |
| What it checks | Whether each claim is accurate | Authorship, plus whether citations exist and are on-topic |
| Method | Cross-checks each claim across 4-5 frontier models | AI-detection plus citation lookup across 220M+ papers |
| Judges if a claim is true? | Yes, that is the whole job | No, by its own FAQ |
| Best moment | Verifying your draft before you publish | Screening for AI writing and missing citations |
| Owned by | Independent verification tool | Superhuman (acquisition announced June 2026) |
The gap: a citation existing is not a claim being true
Picture a sentence that reads: 'A 2024 study found remote workers were 22 percent more productive.' A citation checker confirms the study exists and is about remote work, and passes it. But the study might have said 12 percent, or measured something else entirely, or reached the opposite conclusion. The reference is real; the claim built on it is wrong. That is the majority of what actually damages published work, and it is exactly the layer GPTZero's tools do not cover, as we detail in when AI cites studies that don't exist.
There is a second reason not to lean on GPTZero as a truth signal: it is a detector, and detectors carry documented false-positive risk. GPTZero flagged the US Constitution as AI-written, and a Stanford study found detectors wrongly flagged 61 percent of essays by non-native English speakers. That unreliability is a feature of guessing authorship from style, which is a different question from whether a claim is accurate, a distinction we unpack in why AI detectors ask the wrong question. As of June 2026, GPTZero is being folded into Superhuman Go as an AI-authenticity layer for a writing suite, which is further from a standalone truth-verification workflow, not closer.
Verifying truth needs a different move than detecting authorship or checking that a link resolves. TrueStandard runs the claims in your draft across four to five frontier models from different labs and shows where they disagree; a claim several independent models cannot corroborate is exactly the one to check before you publish. It answers the question GPTZero explicitly declines to.
Pricing and where it's headed
GPTZero is education-first, priced for classrooms and writing teams. Exact rates change and vary by source, so treat these as approximate and check its pricing page:
| Free | Signup required; roughly 10,000 words/month |
| Premium (individual) | Reported around $15-24/mo depending on billing |
| Professional / Team | Reported around $46/mo |
| Enterprise / Classroom | Custom, seat- or volume-based |
GPTZero is priced as a detection and integrity suite for education, and its citation tools ride along with that. TrueStandard is priced as a verification layer for people who publish for a living, where catching one false claim is worth more than the subscription. They price differently because they answer different questions, and because one of them is now an authenticity feature inside a larger writing product.
Which one you need
Use GPTZero if
- — You need to detect whether text was written by AI
- — You are an educator screening student work for fabricated or missing citations
- — You want plagiarism and authorship checks alongside AI detection
Use TrueStandard if
- — You need to know whether your claims are actually true before you publish
- — You want multi-model consensus on accuracy, not a guess about who wrote it
- — Your risk is a wrong statistic or misused source, not just a citation that doesn't exist
Common questions
Does GPTZero detect hallucinations?
It has a Hallucination Detector that flags convincing but non-existent citations, searching over 220 million scholarly articles to catch sources an AI invented. But it checks whether citations exist and are on-topic, not whether your claims are true. GPTZero's own FAQ says the tool does not take a stance on whether claims are true or false.
Is GPTZero a fact-checker?
No, by its own description. GPTZero is an AI-content detector with citation-checking features, and its founder has called the Sources feature 'explicitly not a fact-checker.' It answers whether text was AI-written and whether its citations exist, not whether a claim is accurate.
Can GPTZero tell me if my statistic or claim is correct?
No. Its tools verify citation existence and rough source support, not factual truth. A real citation can still be attached to a wrong number or a misread finding. To confirm a claim is actually correct, you need verification, such as cross-checking it across several independent models, which is a different job from detection.
Did Superhuman acquire GPTZero?
Yes. Superhuman, the Grammarly-descended company now operating under that name (not the standalone email app), announced on June 23, 2026 that it was acquiring GPTZero to fold it into its AI assistant, Superhuman Go, as an AI-authenticity layer. Terms were not disclosed.
Is GPTZero accurate?
As a detector it is widely used, but it carries documented false-positive risk. It flagged the US Constitution as AI-written, and a Stanford study found detectors wrongly flagged 61 percent of non-native English essays. That is another reason authorship detection is not a substitute for verifying whether a claim is true.
Keep reading
AI Detector vs Fact Checker
One asks who wrote this. The other asks is this true. Before you publish, only one of those questions protects your reputation — and most teams are watching the wrong one.
An Originality.AI Alternative
Originality.AI is a strong AI-detection suite. But if the job you care about is verifying claims before you publish, fact-checking is only one of its five bundled checks — and it runs on a single model.
TrueStandard vs Parafact
Both verify claims before you publish. The real difference is what one model can miss — and whether your long-form draft fits inside the check at all.
TrueStandard vs FactCheckTool
These tools look similar and solve opposite problems. One tells you if the media you're consuming is fake. The other tells you if the draft you're about to publish is true.
When AI Cites Studies That Don't Exist
AI does not just get facts wrong. It invents whole sources, cases, studies, DOIs, and cites them with the same confidence it uses for real ones. Here is why it happens, the disasters it has already caused, and how to catch a fabricated citation before your name is on it.
Check whether it's true, not just whether it's cited
GPTZero can tell you a citation exists. It will not tell you the claim is right. Paste your draft into TrueStandard and four to five frontier models check whether each claim actually holds up, in about 60 seconds.
Verify a draft