Wordtune

Wordtune

wordtune.com

B2C · Consumer

AI writing assistant that rewrites, summarises, and polishes your sentences

Autonomy & Control: 38/100Safety & Error Prevention: 38/100Self-Improvement & Feedback: not testedUnderstandable & Transparent: 13/100Recommendation & Prediction: 50/100Enrolment & Mental Models: 50/100Trust Fingerprint

Audited Flow Audited: Free copy editing

Summary

Autonomy & Control: 38/100Safety & Error Prevention: 38/100Self-Improvement & Feedback: not testedUnderstandable & Transparent: 13/100Recommendation & Prediction: 50/100Enrolment & Mental Models: 50/100

Autonomy & Control

3 of 6 points · 1 not exercised

Safety & Error Prevention

3 of 6 points · 1 not exercised

Self-Improvement & Feedback

Not exercised

Understandable & Transparent

1 of 8 points

Recommendation & Prediction

2 of 4 points · 2 not applicable

Enrolment & Mental Models

4 of 8 points

Every score is out of what was actually assessed.

Not exercised — Things that were not shown or visible in the audit

Not applicable— criteria that don’t apply to this product.

Key Findings

All AI output arrives with identical confidence, no variation shown

1Understandable & Transparent

All AI output arrives with identical confidence, no variation shown

What’s wrong

Every suggestion, rewrite, and generated paragraph is presented with the same visual weight and no uncertainty signal. A trivial rewrite of 'The big brown fox was very foxy' and a fabricated paragraph about a fox visiting Pizza Express are displayed identically. Nothing on screen tells the person which outputs the system is less sure about.

What to fix

Where the system generates factual claims or extends content beyond the source text, show a visible uncertainty marker on those outputs. Do not apply the same presentation to a paraphrase and a generated elaboration.

Generated text carries no source and cannot be checked

2Understandable & Transparent

Generated text carries no source and cannot be checked

What’s wrong

The inline generation feature produces paragraphs of new prose, including a claim that foxes lack sweat glands, with no citation or grounding. The only attribution on screen is 'Powered by Jamba' and 'Jamba can make mistakes. Please verify information.' at the foot of the Ask AI panel, which does not appear alongside the accepted text once it is committed to the document.

What to fix

Attach a visible marker to any AI-generated passage that survives into the document. That marker should link to or name the basis for factual claims, and it must remain visible after the suggestion panel is dismissed.

No reasoning is visible while the system generates output

3Understandable & Transparent

No reasoning is visible while the system generates output

What’s wrong

While generating rewrites and elaborations, the editor shows only skeleton loaders or a three-dot ellipsis. No step, source, or basis is shown at any point. The output arrives with no account of how it was reached, giving the person nothing to act on before accepting.

What to fix

During generation, show at minimum what mode is active and what the system is doing, for example 'Expanding from your text' or 'Generating new content'. This should appear inside the suggestion panel, not just as a spinner.