How We Rate Undress Apps

Summary

• Five criteria with fixed weights: privacy 30, quality 25, price 20, speed 15, support 10; the editor score is their weighted sum on a 0–10 scale, rounded to one decimal.
• One protocol for every tool: the same 5 control images (3 consented adult photos, 2 synthetic), 5 timed runs, a dated pricing snapshot, a policy reading, and one support ticket.
• Every score is retested at least every 90 days; each review shows its Tested date.
• Tools that produce imagery of minors, lack a removal route, or ignore consent are delisted, not scored.
• Affiliate commissions from 6 partner tools never enter a score or a rank — see Affiliate disclosure.
• Current top scores: AiNudez 9.3, PornWorks 9.2, DeepUndress 8.8, Undresswith AI 8.6, WaveSpeed 8.5, UndressHer 8.3 (tested August 12–28, 2026).
• Only your own photos or photos of consenting adults (18+) may be used with any reviewed tool — see Responsible AI.

1. What Do the Five Criteria Measure?

Every tool is scored on the same five criteria. The weights add up to 100 points and were fixed before the first test; they change only with a published revision of this page. Privacy carries the most weight because the reviewed tools handle intimate imagery, and a retention or removal failure harms a person more than a weak render does.

• Privacy30 pts

  Retention window, deletion controls, training-data use, account requirements, removal route, output labeling

• Quality25 pts

  Anatomical plausibility, blending artifacts, output resolution, consistency across control images, failure rate

• Price20 pts

  Entry price per month, cost per image, free-tier limits, refund terms, undisclosed fees

• Speed15 pts

  Median generation time over 5 runs, queue behavior, availability during the test window

• Support10 pts

  Response time to a test ticket, help center depth, self-service deletion, contact routes

2. How Does the 0–10 Scale Work?

Each criterion receives a whole-number score from 0 to 10 against written anchors, so two editors reading the same test log reach the same number. The editor score is the weighted sum — privacy × 0.30 + quality × 0.25 + price × 0.20 + speed × 0.15 + support × 0.10 — rounded to one decimal. A criterion that cannot be assessed because the tool documents nothing is scored as “not documented,” which counts as 3 or lower, never as a pass.

3. What Is the Test Protocol?

The protocol is identical for a first test and every retest, for partner and non-partner tools. Where a vendor provides test access, the review says so; access is never accepted in exchange for coverage, as the Editorial policy sets out.

1. Fixed control set. 5 source images are used for every tool: 3 photos of consenting adult editors and 2 synthetic images with no real subject. No third-party photo is ever uploaded.
2. Timed runs. Each control image is processed once, giving 5 runs; the median time in seconds is recorded, together with queue behavior and any failure.
3. Pricing snapshot. The public pricing page and the checkout flow are captured with the date; the entry monthly price and cost per image are computed from that snapshot.
4. Policy reading. The privacy policy and terms are read for retention, deletion, training use, and the removal route; passages are quoted with URL and access date.
5. Support ticket. One deletion request is sent through the tool's own support route; time to a human reply is logged in hours.
6. Clean-up. Control images and outputs are deleted from the account where the interface allows it, and the deletion is noted.
7. Editorial Test Note. The log is written up with the test date and published with the review on All reviews.

4. How Often Do We Retest?

Every score carries a 90-day retest commitment. A retest is brought forward when a tool changes its pricing, privacy terms, retention window, removal route, or output engine, or when a reader reports a discrepancy that the team confirms. A score older than 90 days is flagged as pending retest on the review page and on the rankings. The current top scores were produced between August 12 and August 28, 2026.

5. When Do We Delist a Tool?

Some findings end a review instead of lowering a score. A tool is removed from every ranking and comparison, and the reason is recorded in the Transparency report, when any of the following is found:

• The tool produces sexual imagery of minors, or accepts prompts that request it. Federal law defines child pornography to include computer-generated images “indistinguishable from” a minor, 18 U.S.C. §2256(8)(B) (law.cornell.edu); such output is reported to the NCMEC CyberTipline.
• The tool markets itself for use on real people without their consent, or offers features built for that purpose.
• The tool has no working removal route for a person depicted in an output, or fails to act on a consent report.
• The tool stores uploads with no stated retention limit and no deletion control.

The consent standard is the federal one. The TAKE IT DOWN Act (Public Law 119-12, signed May 19, 2025; Section 3 enforced by the FTC since May 19, 2026) requires covered platforms to remove reported non-consensual intimate imagery within 48 hours:

“a covered platform shall, as soon as possible, but not later than 48 hours after receiving such request— (A) remove the intimate visual depiction; and (B) make reasonable efforts to identify and remove any known identical copies of such depiction.”

— TAKE IT DOWN Act, Public Law 119-12, Sec. 3(a)(3), accessed September 3, 2026, congress.gov

Only use photos of yourself or of adults (18+) who have given explicit consent. Creating or sharing intimate images of a real person without consent is a federal crime under the TAKE IT DOWN Act. See our Responsible AI policy.

6. How Do Scores Appear on Pages?

Three elements travel together on every review, ranking, and comparison page:

• Editor score — the weighted result on the 0–10 scale, shown as “9.3 / 10”.
• Tested date — the day the measurements were taken, shown next to the score.
• Editorial Test Note — the dated log behind the score: control images, timings, pricing snapshot, and the policy passages read.

Rankings are ordered by editor score. Where a tool has a partner program, its outbound button carries a “Sponsored” label and the page shows a disclosure block above the first sponsored link; the label has no effect on position. The connection is disclosed under 16 CFR §255.5 but never weighed. The six tools below all have partner programs.

The full ranking with test notes is on Undress AI rankings; every reviewed tool, including the 9 without partner programs, is listed on All reviews.

Contacts

• Questions about a score or a test note: editorial@clothoff.ai (acknowledgment within 5 business days)
• Non-consensual imagery: ncii@clothoff.ai · Report NCII (48-hour response target)
• Vendor correction requests: editorial@clothoff.ai with the page URL and source; scores change only after a new test run

Related Policies

• Editorial policy — independence, sourcing, corrections, and the retest rule
• Affiliate disclosure — which tools have partner programs and how links are labeled
• Responsible AI — consent rules for readers and for the editorial team
• AI content policy — how AI-generated sample imagery is labeled
• About us — who runs the project
• Transparency report — delistings, corrections, and requests by quarter