Written by people who still have deadlines.

Why this site exists

Most coverage of AI writing tools is either a press release with a byline or a listicle written by someone who opened each tool for ten minutes. Neither tells you what happens on day four, when the novelty is gone and you have a 2,000-word piece due at five.

We started keeping notes on that for ourselves. This site is those notes, tidied up. The name is the whole editorial position: the tools are good at first drafts, and the value you add is the second one.

How we test

Every tool gets the same treatment, so the reviews are comparable with each other:

1. Same brief, every tool. The assignment pack has five jobs: a 2,000-word reported feature from supplied notes, a product landing page, a three-email sales sequence, a fiction chapter with a supplied outline, and a 40-minute meeting transcript to summarise. Each job has a real reader in mind and a real constraint (word count, tone, a fact that must not be invented).
2. A week on real work. After the assignment pack, the tool goes into the reviewer's actual workflow for at least five working days. This is where the friction shows: exports, integrations, limits, what it does when you paste in 9,000 words.
3. Blind read. Output from the assignment pack is stripped of tool names and read by an editor who did not run the test. They score prose quality, factual discipline and how much editing the draft needed before it was publishable. That score is set before anyone looks at the feature list.
4. Price checked, then re-checked. Plans and prices change constantly. Every review and tool card carries the date we last verified them. When a price changes, the review gets a dated update note rather than a silent edit.

Ratings are on a five-point scale and they are editorial: a 4.6 means "we would pay for this and keep using it", a 3.5 means "useful for a specific job, not for most people". They are not averages of sub-scores.

How we make money

Three ways, and all three are visible on the page:

• Affiliate links. When a tool has a partner programme, links to it may earn us a commission if you subscribe. These links are marked and carry a rel="sponsored" attribute. Tools without a programme get reviewed with the same brief and the same rigour — several of our top picks pay us nothing.
• Display advertising. Ad slots are labelled and kept out of the article body. Advertisers do not see reviews before publication and have no say in ratings.
• Newsletter sponsorship, occasionally, marked as such in the email.

What we do not do: paid reviews, sponsored rankings, or "review swaps" with vendors. If a company sends a free licence for testing, the review says so.

Corrections

If a fact, a price or a feature claim is wrong, tell us. Corrections are made in the article with a dated note. We would rather be corrected than quietly wrong.