Detecto
AI Detector │ All modalities

One detector. Four modalities. A report you can defend.

Detecto runs text through a calibrated language-pattern model and runs image, video, and audio through a multi-signal media analyzer. The output is a single report shape (score, signals, notes) across all four.

See a sample reportTry a text scan

Built for the people who review other people’s content

  • Editors at independent magazines
  • University writing centers
  • Newsroom verification desks
  • PR review teams
  • Content QA leads at agencies
Modalities

Pick the modality you actually review.

Each modality page contains the inputs we accept, the signals we expose, an honest limitations block, and a sample on that modality.

Text
Sentence-level highlights, perplexity, burstiness, and a confidence band.
Image
Generation likelihood, metadata integrity, and surface-artifact regions.
Video
Per-scene likelihood timeline, with audio-vs-frame divergence.
Audio
Synthesis likelihood, a confidence timeline, and metadata context.
What’s in every report

Score, signals, evidence. In that order.

Same report shape across all four modalities so reviewers don’t learn a different vocabulary per medium.

Score

A calibrated probability with a confidence band

A single likelihood number plus the variance band so the reviewer sees uncertainty alongside the point estimate.

Signals

Vendor-agnostic feature contributions

Perplexity, burstiness, lexical diversity, and modality-specific features. Each labelled with its contribution toward the verdict.

Privacy

Private by default; export and share opt-in

Reports save privately to your dashboard. PDF export and public share pages are per-scan opt-ins, never default.

Sample report

The same report shape, on a real human passage and a real AI passage.

Two pieces of writing on productivity. One human. One generated.
Full report
Human-writtenrep_h_8b14f3c2
First-person reflection on returning to the office after years of remote work.
Modality: text · May 24, 2026
6%Likely human
0%±4 pts100%
Top signals
  • Average perplexitySurprise per token relative to a base language model.
    74.2
  • BurstinessVariance in sentence-level perplexity across the passage.
    0.81
  • Idiom densityIdiomatic constructions and unconventional phrasing.
    6.4 / 1k
AI-generatedrep_a_2f30d9a1
Short five-paragraph essay weighing the pros and cons of remote work.
Modality: text · May 24, 2026
92%Likely AI-generated
0%±5 pts100%
Top signals
  • Average perplexitySurprise per token relative to a base language model.
    18.3
  • BurstinessVariance in sentence-level perplexity across the passage.
    0.21
  • Connector frequency"First", "Second", "Ultimately", "In today’s world" and similar.
    11.2 / 1k
Open the full sample

Honest limitations

A detection score is probabilistic, not proof. Treat the report as evidence to discuss, not a verdict to publish.

  • A score is a probability, not a verdict. A 78% AI score is the start of a conversation.
  • Confidence widens for very short input. Detecto surfaces this; don’t smooth over it.
  • Generation models change; the calibration plate adjusts on a release cadence.
  • Detection signals can interact with non-native register and creative writing styles.
  • File-based media means we don’t accept URLs or live-stream callbacks in v1.
  • Disagreement between detectors is the rule, not the exception. Pair Detecto with the conversation.
Built for reviewers

Get the report. Make the call.

No card to start. Free tier ships with the full report fidelity.

Get started, freeSee pricing