TruthScan review

TruthScan runs AI detection across text, images, PDFs and documents, voice and video in one product, more content types than any other detector in this guide covers; a few competitors here also check images, but none add documents, voice and video the way TruthScan does. Starter costs $290/yr, shown on the pricing page as an annual-billed rate of about $24/mo, for 500,000 words of text and 1,000 image or PDF checks a month, working out to roughly $4.83 per 100,000 words and about 2.4 cents per image or document check. The real evidence problem shows up exactly where TruthScan markets itself hardest: an independent test of its ID-verification feature found a 57% false-positive rate on legitimate IDs and missed 29% of AI-generated fakes, against the vendor’s own 99% headline claim. G2 rates it 4.8 out of 5, but from just 5 reviews as of October 2026, too thin a sample to lean on either way.

What TruthScan is for

TruthScan positions itself at fraud, trust-and-safety and compliance teams, banking, insurance, legal, HR and media, rather than the academic-integrity or editorial audience most of this category serves. Text detection runs as its own product line with its own API documentation and its own word-credit pricing, not a minor add-on bolted onto the image side. Image detection returns a confidence score, a plain-language explanation and a region-level heatmap; document detection targets receipts, invoices and verification PDFs for metadata, font-consistency and edit-history red flags, aimed squarely at refund and application fraud. Voice and video detectors exist as named products too, though we couldn’t independently confirm separate pricing for either beyond the general credit system. One disclosure worth knowing before trusting any of this: TruthScan’s co-founder and CEO, Christian Perry, also co-founded Undetectable.ai, an AI humanizer built to help text evade detectors like this one, according to Forbes’ own profile of Perry (forbes.com/profile/christian-perry, 2026-10-31).

What you get on each plan

One “result” is one image or one PDF page; text runs on a separate one-credit-per-word meter. On Starter, that works out to about $4.83 per 100,000 words of text ($24.17 ÷ 500,000 × 100,000) and about $0.024 per image or document check ($24.17 ÷ 1,000), both on the annual-billed rate; the page doesn’t list a month-to-month price for Starter, though the same “2 months free” math TruthScan confirms for Business would put it near $29/mo. Overage runs $0.03 per extra result or per 1,000 extra words on Starter, dropping to $0.01 per result on Business; text overage past Starter isn’t published.

Accuracy: strong on paper, uneven in the one independent test that exists

TruthScan’s own marketing, and a sponsored VentureBeat contributor piece, claim accuracy above 97% for photorealistic deepfakes and a general 99% headline figure, plus a self-reported 99.3% on AI-edited refund-fraud images specifically. Those are vendor numbers, not test results. The one independent, methodology-disclosed test, IDScan.net’s structured comparison of AI-generated IDs, legitimate IDs and traditional counterfeits, all photographed under matching conditions, tells a different story: TruthScan correctly flagged 71% of AI-generated fake IDs, missing the other 29%, and incorrectly flagged 57% of legitimate IDs as fraudulent. IDScan.net sells a competing document-authentication product, so read this as a competitor-run test with a disclosed methodology rather than a neutral academic study, but it’s still more rigorous than most vendor marketing in this category, and ID verification is one of TruthScan’s own named use cases, not a stretch comparison. On the text side, an independent CyberNews nine-sample test measured 65% overall accuracy, including a 79% AI score on a 50/50 human-AI hybrid sample and a 27% false-positive score on a fully human sample. A separate 62% false-positive figure for the text detector circulates in secondary coverage, attributed to “TwainGPT,” itself a competing detector and humanizer brand; that’s a competitor-sourced claim, flagged here rather than reported as fact. TruthScan’s accuracy varies by what you’re checking, and the gap between its own 97-99% claims and the one independent test available is wide enough to treat as the headline finding, not a footnote.

What users say in reviews

Every platform here is thin. G2 rates TruthScan 4.8 out of 5 from 5 reviews, and a Chrome Web Store listing for its image detector rates 5.0 out of 5 from 24 ratings and roughly 2,000 users. Trustpilot shows a 3.2-star average, but it comes from exactly one review, posted by a user complaining about an unpaid link-building services invoice, not about detection accuracy; Trustpilot itself flags the profile as unclaimed with reviews that “may not be representative,” so we’re not treating that score as a real signal either way. Capterra’s listing exists but couldn’t be independently confirmed, and it shows a stale $49/mo starting price that doesn’t match the live pricing page. The closest thing to a recurring theme in the thin sample that does exist: a G2 reviewer noted the pricing “didn’t make sense for a small retail store that ran only occasional checks,” a volume-fit complaint rather than a product-quality one, alongside praise for the detection heatmap and self-service access.

Who should skip it

TruthScan is a weak fit for anyone checking only written text for AI authorship; GPTZero, Originality.ai and Quetext all serve that academic and editorial use case more directly, with larger review bases to judge them by. It’s also a weak fit for anyone leaning on the ID-verification feature specifically without reading the IDScan.net test first, given the gap between that 57% false-positive finding and the vendor’s own 99% claim. And anyone weighing TruthScan’s accuracy pitch against its CEO’s other company, a humanizer built to beat detectors like this one, should read that disclosure before trusting either company’s numbers uncritically. See the full comparison, including Copyleaks and Winston AI, in Best AI Detectors.