AI Writing Feedback Examples: From Draft to Revision
Below are three complete WriteFit training rounds. Each shows the prompt, the writer's draft, three diagnoses that each quote the draft, the writer's own revision, a note on what changed, and the goal for the next round. The AI only diagnoses; the revision is the writer's work.
Note: all drafts are written by the WriteFit team, not users. In Examples 1 and 2 the diagnosis is the live product's real, unedited output; every revision, and the diagnosis in Example 3, is hand-written. None of this is user data or an efficacy claim. To see real output, run your own draft through the no-signup demo. Try the demo
Example 1: Founder product intro
Source: draft written by the WriteFit team; the diagnosis below is the live product's real output, unedited (run 2026-10-25). The revision is written by hand.
Prompt
In about 80 words, explain what your product does and who it is for, to someone who has never heard of it.
Original draft
Our platform leverages cutting-edge AI to empower teams to unlock their full potential. It is a seamless, all-in-one solution that streamlines workflows and drives productivity. Whether you are a startup or an enterprise, our product is designed to meet your needs and help you achieve your goals.
AI diagnosis: 3 specific issues
1. Buzzword stacking in place of a concrete claim
Quoted: leverages cutting-edge AI to empower teams to unlock their full potential
Why it matters: 'Leverages', 'cutting-edge', 'empower', and 'full potential' are four abstractions in a row. A reader cannot picture a single thing your product does, so there is nothing to remember or repeat. This exact phrase could sit in a thousand other landing pages.
Revision task: Delete all four buzzwords. Then write one sentence naming a specific person doing a specific task: who they are, what they are doing today that is slow, and what your product does instead. No word 'solution', 'empower', or 'leverage' allowed.
2. Claims with no measurable referent
Quoted: seamless, all-in-one solution that streamlines workflows and drives productivity
Why it matters: 'Streamlines workflows' and 'drives productivity' are unfalsifiable. The reader cannot check them, so they discount them. 'Seamless' and 'all-in-one' are adjectives the reader has been trained by a decade of SaaS copy to ignore.
Revision task: Replace each of these three claims with a number or an observable behaviour: how many steps removed, how many minutes saved, which tools are replaced. If you don't have the number, describe the before-and-after scene instead.
3. Empty audience claim that addresses everyone
Quoted: Whether you are a startup or an enterprise, our product is designed to meet your needs and help you achieve your goals.
Why it matters: A claim that covers every customer tells each reader that you were not thinking of them. 'Designed to meet your needs' and 'achieve your goals' are placeholders where the actual benefit should be. This sentence creates resistance rather than trust.
Revision task: Cut the startup-or-enterprise hedge. Name the one customer you serve best, and state what they get that a competitor's customer does not. Then delete the words 'needs' and 'goals' from the sentence entirely.
The writer's own revision
Our tool turns a recorded customer call into a three-line summary and a to-do list for the account manager, within a minute of hanging up. It is built for small B2B sales teams who currently type call notes by hand after every meeting.
What changed
The writer cut every abstract adjective, named one user (small B2B sales teams) and one painful habit (typing notes by hand), and described input and output. The revision is 47 words — shorter, yet the reader now knows what happens.
Next goal: add one number or one concrete before/after (for example, minutes saved per call) without adding a stock phrase.
Example 2: Personal narrative
Source: draft written by the WriteFit team; the diagnosis below is the live product's real output, unedited (run 2026-10-25). The revision is written by hand.
Prompt
Write about a moment you realised you had changed. Show it through one scene, not a summary.
Original draft
Moving abroad changed me a lot. I used to be shy, but over time I became much more confident. It was a difficult journey full of challenges, but I learned so much about myself and grew as a person.
AI diagnosis: 3 specific issues
1. Summary statements instead of scenes
Quoted: Moving abroad changed me a lot.
Why it matters: This is a verdict, not an experience. The reader is told the conclusion before seeing any evidence, so there is nothing to picture and nothing to believe. Any person who has ever moved abroad could have written this sentence.
Revision task: Delete this opening line and replace it with one specific moment that shows the change happening — a time, a place, and what you did or said. Write the scene first; let the conclusion come at the end, or not at all.
2. Cliché framing of difficulty
Quoted: It was a difficult journey full of challenges
Why it matters: 'Journey' plus 'full of challenges' is a phrase so worn down it carries no information. The reader cannot feel difficulty in the abstract; they feel it in a specific failure — a call you couldn't make, a form you couldn't fill in, a dinner you sat through in silence.
Revision task: Name one actual hard moment. What was the situation, what went wrong, and what did you do in the next ten minutes? Replace the whole clause with that single incident.
3. Empty self-assessment ending
Quoted: I learned so much about myself and grew as a person.
Why it matters: This is the sentence every personal essay ends with, and it means nothing because 'so much' and 'grew' are unmeasurable. It also asks the reader to trust your self-report rather than your evidence.
Revision task: Cut the sentence. Then write one short line of evidence for growth — a specific behavior you now do that the old you could not have done (a phone call you make, a room you walk into, a question you ask). If you can't name one, the essay isn't finished yet.
The writer's own revision
In my second month in Toronto, a barista asked for my name and I said it the way my mother does, without apologising for the vowels. She wrote it down correctly on the first try. I carried the cup to the window and noticed I was not rehearsing the order in my head any more.
What changed
The change is now shown by one action (saying the name without apologising) and one small observation (no longer rehearsing). There is no claim about confidence at all; the reader infers it.
Next goal: add one line of dialogue that carries tension, without explaining what the narrator feels.
Example 3: Opinion and argument
Source: draft, diagnosis and revision all written by the WriteFit team as an illustration.
Prompt
Argue for or against: 'Remote work makes teams less creative.' Take a side in about 100 words.
Original draft
Remote work has both advantages and disadvantages for creativity. On the one hand, it offers flexibility. On the other hand, it reduces spontaneous interaction. In conclusion, it depends on the team and the situation, and companies should find a balance that works for them.
AI diagnosis: 3 specific issues
1. No position is taken
Quoted: "it depends on the team and the situation"
Why it matters: The prompt asked you to pick a side. A balanced non-answer is safe for the writer but leaves the reader with nothing to agree or disagree with.
Revision task: Write one sentence beginning 'I think remote work ... because ...' and make it the first line.
2. Claims without evidence
Quoted: "it reduces spontaneous interaction"
Why it matters: Which interaction, and what creative output was lost? Without an example the claim cannot be tested.
Revision task: Give one concrete moment (a hallway chat, a whiteboard) and say what idea came out of it.
3. Template structure
Quoted: "On the one hand ... On the other hand ... In conclusion"
Why it matters: The three-part scaffold is the shape of an exam answer; it hides whether you actually have an argument.
Revision task: Drop the scaffold. Lead with your claim, give one example, then answer the strongest objection.
The writer's own revision
Remote work does not make teams less creative; it makes them less accidentally creative, which is fixable. My last team's best idea came from a hallway comment about a customer email. After we went remote we lost those comments, so we started a ten-minute 'half-baked ideas' call every Friday. Within a month it produced more pitches than the office ever did.
What changed
The writer commits to a position in sentence one, replaces 'reduces interaction' with a specific lost moment, and answers the objection with an action taken. The on-the-one-hand scaffold is gone.
Next goal: add the strongest opposing view in its own sentence, stated fairly, before you rebut it.
Who WriteFit is for, and who it is not for
Good fit: people who want to write more clearly and specifically in English or Chinese — founders, bloggers, students, professionals — and are willing to revise their own text.
Good fit: non-native English writers who want to keep their own voice while fixing clarity and specificity.
Not a fit: people who want the AI to write or rewrite for them (WriteFit never does), or who need exam scoring such as IELTS or TOEFL band estimates.
Not a fit: AI-text detection. The 'AI-likeness' score flags habits such as stock phrases and symmetrical structures. It does not and cannot prove who wrote a text.
English and Chinese. You can write in either, and the feedback follows the language of the interface you choose.
What kinds of writing can I practise?
Free writing, persuasive and argument pieces, narrative and personal stories, short founder or product descriptions, and sentence-level rewrites. It is not built for long manuscripts or academic papers.
What level is it for?
Beginner to advanced writers. The diagnosis works on whatever you submit, but it is tuned for short pieces of roughly 50 to 300 words, not full essays.
Does WriteFit give CEFR levels or exam scores?
No. It gives three-band judgements on clarity, specificity, voice and AI-likeness. These are an editor's judgement, not a validated exam score.
Can the AI feedback be wrong?
Yes. It is a language model. Each issue quotes your text so you can check it yourself, and you decide whether to accept it. If you disagree, keep your version.
Is there proof that it improves writing?
Not yet in the form of a controlled study. The method is based on deliberate practice and revision, which is well studied, but we do not claim measured outcomes for WriteFit users yet.