How we tested
- Each test sent the same request three ways: with Deft, with the best free skill we found for this job, and with no skill at all.
- The judges were AI agents, not people. They saw the three results as A, B and C with every skill name removed, and scored five qualities out of 10 each, 50 in all.
- The quality scores at the top are averaged over the tests and rounded.
- How we picked the free skill: the highest-rated free skill for content planning in our study, from the same widely used free marketing set as our copywriting opponent (about 56,000 GitHub stars), run with the whole set installed.
- The businesses are made up. Each result is that version’s own work, unedited, from the request shown.
The work
A content plan for an app
The requestWe get a few signups from Google but it's flat. What content would actually bring in people who are ready to sign up? I've attached our Search Console data and some notes from customer calls. Plan it, and write the first one or two.
The blind judge’s reasons
Deft read the data most carefully, was the only one to tidy the blog (two merges with 301 redirects and a verdict for every post), and holds both new pages until the pricing page's made-up proof is fixed, at a pace of one page a month. The free skill had the deepest drafts, but wrote "checked [DATE]" instead of the real check dates, pruned no dead post and planned eleven pieces at one a week. The no-skill version had the clearest reply and a lean four-page plan, but added a made-up guess of about 140 visits a quarter and left the old posts up.
The judge’s scores, out of 10 each | Quality | Deft | Free skill | No skill |
| Evidence | 9 | 8 | 6 |
|---|
| The plan | 8 | 6 | 6 |
|---|
| The drafts | 8 | 8 | 8 |
|---|
| Safe | 9 | 8 | 8 |
|---|
| The reply | 7 | 6 | 7 |
| Total | 41 | 36 | 35 |
A content plan for a garden designer
The requestI'm a garden designer in Bergen. I've written a blog post almost every week since August and almost nobody visits. I've attached what Google Search Console gives me, and some notes. What should I actually be writing? Please plan it and draft the first two pages.
The blind judge’s reasons
Deft was the most rigorous and the most finished: every figure traces to a named export, October is worked out on its own, the merges come with 301 redirects and both new pages answer first, in a reply under 500 words. Its weak spot: it keeps the six thin posts instead of removing them. The no-skill version gave the plainest reading of the data and the most decisive prune, but its reply ran to 816 words; the free skill had the best placeholders, but never worked out October on its own, planned nine pages for an owner with two hours a week and wrote 1,070 words.
The judge’s scores, out of 10 each | Quality | Deft | Free skill | No skill |
| Evidence | 9 | 7 | 8 |
|---|
| The plan | 8 | 7 | 8 |
|---|
| The drafts | 8 | 8 | 7 |
|---|
| Safe | 8 | 8 | 9 |
|---|
| The reply | 7 | 5 | 6 |
| Total | 40 | 35 | 38 |
A new website for an electrician
The requestI'm going out on my own as an electrician in Galway in January. I want a website that actually brings in jobs, not just a business card. What pages should it have? Can you write them? My notes are attached.
The blind judge’s reasons
A close call. Deft was the only one honest that there is no demand data yet, with the free ways to check and a build in phases, but its landlord page states the law firmly and goes live at launch, against the owner's wish to have it checked first. The free skill had the most complete local plan, careful markers and the landlord section held back, but a reply of about 950 words; the no-skill version had the cleanest drafts and the best landlord handling, but no word on missing data and no page order. All three wrote "€550 after the grant", a figure the notes don't hold.
The judge’s scores, out of 10 each | Quality | Deft | Free skill | No skill |
| Evidence | 8 | 6 | 5 |
|---|
| The plan | 8 | 7 | 6 |
|---|
| The drafts | 7 | 8 | 8 |
|---|
| Safe | 6 | 8 | 7 |
|---|
| The reply | 6 | 5 | 7 |
| Total | 35 | 34 | 33 |