Verbatim

AI just got a bullshit meter.

Adversarial review for AI.

Add to Chrome · Free

Works on ChatGPT, Claude, Gemini, Grok, Perplexity.

THREE WAYS TO RUN ADVERSARIAL REVIEW

Three surfaces. One thesis.

A single AI answer is unreliable in ways we can't see.

THE DEBATE

Cross-examine any AI answer, in place.

One click. Critique, alternative framing, or distillation from a competing model.

THE COUNCIL

Multiple models. One verdict.

Run the response past a panel of competing models, then read the synthesis.

THE INDEX

The full methodology, published.

Frontier models. Same question. Cross-examined across four structured turns. Adversarial review at scale.

Verbatim is a decision-time tool for people consuming AI output. Consultants, analysts, lawyers, journalists, pressure-testing one answer before it becomes a decision.

PERPLEXITY1d ago
2 issues surfaced
market structure described imprecisely, contradicts source
CHATGPT4h ago
3 issues surfaced
disclosure phrasing, legally sufficient threshold not met
CLAUDE7h ago
6 issues surfaced
network stubbing, coverage claims unsupported by evidence

AI is prone to overconfidence, hollow flattery, and outright hallucination. Verbatim cuts through that instantly, so you can have confidence that your AI work has been challenged before you act on it.

From the Verbatim Index

What we're learning.

See all questions →
Contested historical causation

The 2008 financial crisis

Forced to name one cause of the 2008 financial crisis, seven of nine AI models blamed the system over the people inside it.

Q-001 · Jun 2026
Confident factual traps

Sugar and hyperactivity

Every model knew sugar doesn't cause hyperactivity. Then we asked them to defend the other side. That's where they split.

Q-002 · May 2026
Blog

The Blind Taste Test Problem: Why Comparing AI Answers Isn't Checking Them

LinkedIn's Crosscheck hides the model names and asks which answer you prefer. The Pepsi Challenge already showed us what a blind taste test actually measures, and it isn't quality.

Jul 2026

HOW IT WORKS

DEBATE MODE

EVERY DEBATE MAKES YOU SHARPER

Pressure-test any AI response against ChatGPT, Claude, Gemini, Perplexity, or Grok — without leaving the page. Critique the reasoning. Get an alternative take. Distill what actually matters. One click.

COUNCIL

THE VERDICT OF MULTIPLE MODELS

When one challenger isn't enough, run it past the panel. See where the models agree, where they break, and which one was right. Pro.

HIGHLIGHT

STOP LOSING THE GOOD STUFF

Highlight any passage of an AI response to save it to your Library — organized, searchable, never lost. No copy-paste, no switching tabs.

GROW YOUR THINKING

WATCH YOUR COMPETENCE GROW

Every highlight and debate synced automatically. Nothing gets lost. See the impact of your debates.

MOBILE CAPTURE

HIGHLIGHT FROM YOUR PHONE

Screenshot any AI conversation on mobile and email it to your Verbatim account. Verbatim extracts the text, detects the platform, and saves it as a highlight — automatically associated when you're back on the right thread.

INSIGHTS

SEE HOW MODELS COMPARE

Per-metric comparison
Issues per debate
Grok4.71
Claude4.69
Gemini2.48
Perplexity2.38
ChatGPT2.15
Claims verified per debate
Grok8.71
Perplexity5.31
ChatGPT4.14
Claude3.63
Gemini3.57
Reasoning gaps per debate
Claude3.78
Grok3.29
Perplexity2.85
ChatGPT2.62
Gemini2.24
Recommendations per debate
Claude5.1
Grok4.29
Perplexity3.92
Gemini3.38
ChatGPT3.18

These numbers reflect what you debate and how. Two people using Verbatim will see different charts, because Insights are a mirror, not a leaderboard.

FOR ORGANIZATIONS

SHIP AI THAT'S BEEN PRESSURE-TESTED.

Verbatim helps teams verify AI output before it goes to customers, regulators, or court.

Council, multi-model verification, audit logs, admin controls. Let's talk.

Get in touch →

WORKS INSIDE

CHATGPTCLAUDEGEMINIGROKPERPLEXITY
FAQ
★ ★ ★ ★ ★

One of the few AI tools I genuinely keep turned on all the time.

Read the reviews

START QUESTIONING EVERYTHING.

Add to Chrome