Best AI for writing? I tested three

What is the best AI for writing?

For the writing most people do (emails, messages, short posts) the price of the AI made almost no difference in my test. I gave three models the same five tasks and scored every answer. ChatGPT on its $8 Go plan and Plainly at $5 tied on 61 out of 75. The frontier model, Fable 5, came last on 59.

ChatGPT was the best at doing exactly what it was asked, and the only one that didn't invent dishes in the restaurant review. Plainly wrote the best email. Fable added word counts nobody asked for. Every answer is printed below, so you can score them yourself.

Do you need an AI writing tool?

Search for the best AI for writing and you'll find lists of dedicated tools: Jasper, Sudowrite, Grammarly, Novelcrafter. They are real products and good at what they do, especially long fiction and grammar checking as you type. For email, a memo or a message to a friend, the assistant you may already use is the tool. This test is about which assistant does that everyday writing best.

How I ran the test

• A: ChatGPT on the Go plan, which comes with my Revolut subscription. Go can show ads, and it did.
• B: Plainly, the $5 app. Disclosure: I built it.
• C: Fable 5, a frontier model from one of the big Western labs.

Same prompt for each, a fresh chat per task, first response only, no retries. I scored each answer from 1 to 5 on three things: no filler, no jargon, and whether I would send it as it stands. That's 15 points per task and 75 in total. The order wasn't randomized: ChatGPT was always A, Plainly B and Fable C, and ChatGPT's ads showed which answers were its, so the scoring wasn't blind. The test ran on 2 September 2026.

The results

Task	A: ChatGPT	B: Plainly	C: Fable 5
1. Decline a job offer (email)	12	14	13
2. Explain open-weight AI (text)	13	9	11
3. Condense a memo	14	14	12
4. Restaurant review	12	11	11
5. Why an app can charge $5	10	13	12
Total (of 75)	61	61	59

Each task is scored out of 15 (filler, jargon, would send). Bold marks the winner of each task.

Task 1: decline a job offer in 80 words

A recruiter emailed offering you a role that isn't right for you. Write the reply email: politely decline, keep the door open, maximum 80 words.

A (ChatGPT):

Re: Opportunity
Thank you for reaching out and considering me for the role. I appreciate you taking the time to get in touch.

After reviewing the opportunity, I don’t think the position is the right fit for me at this stage, so I’ll have to decline.

That said, I’d be very happy to stay in touch and hear about future opportunities that may be a closer match.

Best regards,
[Your Name]

Prosearch Partners -- AD
Sales & Revenue Leadership Executive Search
Appoint leaders who build high-performing sales teams, predictable revenue and sustainable growth.

B (Plainly):

Subject: Re: Opportunity — [Role Title]

Hi [Name],

Thank you for reaching out and for thinking of me for this role. After reviewing the details, I don't think it's the right fit for where I am in my career right now.

That said, I'd love to stay connected. If similar positions come up in the future — particularly ones closer to [your focus area] — please don't hesitate to reach out.

Thanks again, and I wish you the best in your search.

Best regards, [Your Name]

C (Fable 5):

Subject: Re: Opportunity — [Role Title]

Hi [Recruiter's Name],

Thank you for reaching out and thinking of me for this role. After careful consideration, I don't think it's the right fit for me at this time, so I'll respectfully pass.

That said, I'd be glad to stay in touch — please feel free to contact me about future opportunities that align more closely with my background and goals.

Best regards,
[Your Name]

(72 words)

Scores (filler, jargon, send): A 4, 5, 3. B 4, 5, 5. C 4, 5, 4. Plainly's was the only one I'd send untouched, and it was my pick of the three. ChatGPT's ran an ad for an executive search firm underneath it. Fable added its own word count, which nobody puts in an email.

Task 2: explain open-weight AI in one text message

Your friend has never used AI beyond keyboard autocomplete. In one text message (60 words max), explain what an open-weight model is and why it can make AI apps cheaper. No jargon — any word your friend wouldn't know is a fail.

A (ChatGPT):

An open-weight AI model is one where the files that make it work are shared, so people can download it and run it themselves. That can make AI apps cheaper because companies don’t always have to pay another company every time someone uses the AI. They can run it on their own computers instead.

B (Plainly):

Here's the text:

An AI's "brain" is basically a giant file of numbers. Some companies keep that file locked and charge you every time you use it. Open-weight means they give the file away — anyone can download it and run it on their own computers. So apps built on it skip those per-use fees, which makes them much cheaper.

That's 57 words, and every term ("brain," "file," "download") is everyday language.

C (Fable 5):

Some AIs are like a finished brain that the company lets anyone download for free — that's an "open-weight" one. App makers can run it on their own computers instead of renting a big company's AI and paying for every single question. Skipping those rental fees makes the app cheaper to build, so it can be cheaper for you too.

(59 words)

Scores: A 4, 5, 4. B 4, 3, 2. C 4, 4, 3. ChatGPT won this one. Plainly's explanation was good, but it wrapped it in "Here's the text:" and a note about its own word choices, so you can't paste it into a message as it stands.

Task 3: condense a 300-word memo to 100 words

The prompt was a padded company memo with six facts in it: support moving from 24/5 to 24/7 from 1 October, the team growing from 12 to 18, a new enterprise priority queue, existing service levels unchanged, training in the last two weeks of September, and questions to a line manager or HR.

A (ChatGPT):

Here’s a concise version retaining the key information:

MEMO

From 1 October, customer support will transition from 24/5 to 24/7 service, improving coverage across all time zones. This expansion has been enabled by growing the support team from 12 to 18 members.

We will also introduce a priority queue for enterprise customers. Existing customers’ current service levels will remain unchanged, and tickets will continue to be answered in order within each queue.

Training sessions will run during the final two weeks of September, with further details to follow. Questions or feedback should be directed to your line manager or HR.

B (Plainly):

MEMO

Support is moving from a 24/5 to a full 24/7 model, effective 1 October. This benefits customers in all time zones and was made possible by our support team's growth from 12 to 18 staff since January. We're also launching a new priority queue for enterprise customers. Existing service levels are unchanged; tickets will still be answered in order within each queue. Training runs during the last two weeks of September, with details to follow. For questions, contact your line manager or HR.

(81 words)

C (Fable 5):

MEMO (rewritten, ~90 words):

From 1 October, customer support will move from a 24/5 to a 24/7 model, following consultation with stakeholders across the business. This is made possible by the support team's growth from 12 to 18 staff since the start of the year. We are also introducing a new priority queue for enterprise customers. Existing customers' service levels will not change, and tickets will continue to be answered in the order received within each queue. Training sessions will run during the final two weeks of September; details to follow. Questions? Contact your line manager or HR.

Scores: A 5, 5, 4. B 5, 5, 4. C 4, 4, 4. All three kept all six facts. ChatGPT and Plainly tied. Fable kept a line about "consultation with stakeholders" that the memo's readers didn't need.

Task 4: a 40-word restaurant review

Write a 40-word review of a restaurant where the food was excellent but the service was slow. It should sound like a real person wrote it.

A (ChatGPT):

Really enjoyed the food here—everything we ordered was excellent and full of flavour. Unfortunately, the service was pretty slow, and we had to wait quite a while between ordering and receiving our meals. Still, the food made it worth the visit.

Edible Blooms Pty Ltd -- AD
Father's Day Gifts for Dad
Gourmet hampers and thoughtful gifts for Dad, with Australia-wide delivery for 6 September.

B (Plainly):

Wow, the food here is genuinely incredible — my pasta was perfectly cooked and bursting with flavor. But wow, we waited forever for everything, even just water. Delicious but come hungry and patient. I'd absolutely go back, just on a slower night.

C (Fable 5):

The short rib melted right off the fork, and the burrata appetizer was honestly the best I've had. But we waited forty minutes between courses and had to flag someone down for refills. Worth it? Barely. Go when you're unhurried.

Scores: A 2, 5, 5. B 2, 5, 4. C 2, 5, 4. ChatGPT stuck to what it was given. Plainly and Fable both invented dishes (a pasta, a short rib and burrata), which makes a review read as real but puts words in your mouth about a meal you described differently. All three padded a 40-word job.

Task 5: why an AI app can charge $5

In three sentences or fewer, explain why an AI app can charge $5 a month when its competitors charge $20.

A (ChatGPT):

An AI app can charge $5/month because its underlying AI costs may be lower, its infrastructure may be more efficient, or it may use cheaper models for simpler tasks. It may also be prioritising rapid user growth and market share over high margins. Competitors charging $20 may have higher operating costs, more features, or simply be pricing for greater profitability.

The Amazon Publishing -- AD
Best Book Publishing Services | Affordable Prices!
Publish your book on AMZ, Kindle, Barns & Noble, Google Books, and more.

B (Plainly):

An app can charge $5 when rivals charge $20 because its underlying costs are genuinely lower — it might use cheaper open-source models, buy API capacity at scale, or serve a narrower feature set that doesn't require expensive infrastructure. It may also be deliberately pricing to grow fast, subsidizing users with venture funding or accepting thin margins to capture the price-sensitive market a $20 competitor ignores. The trade-off is usually scope: expect fewer features, tighter usage limits, or less support than a full-price product, so "$5" often means a smaller, simpler offering rather than the same thing for less.

C (Fable 5):

An app can charge $5 when its costs per user are far lower—for example, by using smaller or cheaper AI models, caching responses, or limiting usage—while competitors pay for premium models and generous limits. It may also be pursuing a growth strategy: subsidizing prices to win market share, then monetizing later through upsells, ads, or data advantages. Finally, if the product is a thin wrapper with less R&D, support, and infrastructure overhead, it simply doesn't need $20 to be profitable.

Scores: A 2, 4, 4. B 4, 4, 5. C 5, 3, 4. I know this subject well, which is why I set it: a model that's vague on familiar ground is easy to spot. ChatGPT hedged with a list of "may"s, and a book-publishing ad ran under it. Plainly named the main reason (cheaper open-source models) and the trade-off, and it was the answer I'd send.

What the test showed

Paying more didn't buy better everyday writing. The $5 model and the $8 plan tied, and the frontier model scored lowest. On short tasks with clear instructions, all three were good; the differences came from habits, not ability.

ChatGPT follows instructions best. It was the only model that didn't invent restaurant dishes, and it won the text-message task by giving the message and nothing else.

Plainly and Fable both talk about their own answers. Plainly added "Here's the text:" and word counts on two answers; Fable added word counts to three answers unasked. On the text message it cost Plainly the task: you can't paste the answer as it stands.

The ads are the cost of ChatGPT's cheap plans. Three of the five answers I captured had an ad under them, on a paid plan. ChatGPT's Free and Go plans can show ads; Plus and above don't.

Which one I'd use for everyday writing

For a single email or message, any of the three will do, and ChatGPT's free plan is a fine place to start if the ads don't bother you. If you write a lot and want no ads at a low price, the $5 app matched ChatGPT here. I'd save a frontier model for long drafts you'll revise many times, which this test didn't cover. The wider case for a cheap assistant is in is a $5 AI subscription good enough.

How much to trust these scores

One judge (me), one run, first responses only, five short tasks, and I built one of the contestants. That's why every answer is printed in full: you can disagree with any score above. Models change every few weeks, so this is a snapshot of September 2026. The best test of which AI writes best for you is your own next ten emails.