Skip to content
reebuff
Research

We read all 15,852 posts on DeepAI’s feedback board. Here is what users complain about

Before comparing DeepAI Code with Freebuff, we wanted to know what DeepAI’s own users say. Most of the board is spam. The rest is about paying, quality and things that break.

In short

  • DeepAI’s public feedback board held 15,852 posts on November 7, 2026. About 89% were automated spam floods. That left 1,690 genuine, unique posts.
  • Two thirds of the genuine posts (1,129) are complaints. Only 64 (under 4%) are praise.
  • Paying is what users care about most: 7 of the 10 most-upvoted posts ask DeepAI to make it free or stop the upgrade prompts, and an eighth asks for unlimited chat. Another 122 posts are mainly about billing: charges, refunds and cancelling.
  • The most common complaint is output quality (388 posts), mostly images that ignore the prompt. Then come bugs (276) and content filters that flag harmless requests (121).
  • Almost all of it concerns DeepAI’s chat, image and character products. DeepAI Code is only weeks old: so far the board has two sign-in failures and a request for a Linux version.

DeepAI now has a coding agent, DeepAI Code, that competes directly with Freebuff. Comparing feature lists only tells you so much, so we went to the source: DeepAI’s public feedback board at deepai.org/feedback, where anyone can post a bug, request or complaint and other users upvote it. We downloaded every post on the board and read through them.

How we did it

On November 7, 2026 we collected all 15,852 posts through the board’s own public listing. Five automated floods made up 14,036 of them: thousands of copies of a Discord invite, walls of repeated characters, spam links and the like. After removing those and exact duplicates, 1,690 genuine posts remained.

We labeled each genuine post with its main theme, such as output quality, bugs, billing or a feature request, using AI-assisted labeling that we spot-checked by hand. Treat the counts as close estimates, not exact figures. The board does not show dates, and it covers every DeepAI product, so most posts are about the image generator (597), AI chat (421), characters (85) and video (69).

The numbers

  • Output quality: 388 posts (23%)
  • Bugs and errors: 276 (16%)
  • Feature requests: 271 (16%)
  • Billing, refunds and cancelling: 122 (7%)
  • Content filters: 121 (7%)
  • Paywall and upgrade prompts: 106 (6%)
  • Praise: 64 (4%)
  • Usage limits: 51 (3%)
  • Lost chat history: 43 (3%)
  • Ads: 41 (2%)
  • Account and sign-in problems: 37 (2%)

Counting posts where a theme came up at all, not only as the main point, billing appears in 184 posts, the paywall in 166, usage limits in 112 and lost history in 108.

1. Paying is what people upvote

Paywall posts are only 6% of the board, but they collected 29% of all upvotes (282 of 973). The most-upvoted post on the whole board, at 53 votes, is a plea in capital letters to "STOP WASTING OUR TIME AND MONEY PUTTING UP THE SUBSCRIPTION AD." Six more posts in the top ten say the same thing in fewer words: make it free, stop the upgrade pop-up, why did you add pricing.

A recurring version of this complaint is about things that used to be free moving behind Pro or a sign-in. Recent posts protest a chat model that moved to Pro and a change that made people sign in to use chat models they had used without an account.

Billing is the other half. Users say they cannot find the button to unsubscribe, were charged after cancelling, paid for Pro and still hit limits, or lost credits on generations that failed. DeepAI’s subscribe dialog says Pro is non-refundable.

2. Quality: outputs that ignore the prompt

Output quality is the most common main complaint, and 271 of those 388 posts are about the image generator. The patterns repeat: negative instructions that produce the opposite (more than a dozen posts say asking for "no beard" or "clean shaven" adds a beard), garbled text inside images, scenes that ignore what was asked, and edits that return the image unchanged but still use up credits. On the chat side, users report answers that stop mid-sentence and a paid mode that got worse at following simple instructions.

3. Things that break

The bug reports cluster too. About 15 recent posts describe characters locking users out with "This character doesn’t want to talk to you right now." Others report errors when creating a character, video generation stuck on a network error, voice chat that cannot hear the microphone, and a login link that does nothing.

4. Filters, history and ads

  • Content filters (121 posts): harmless prompts such as family photos, a crowd or a wrestling pose blocked as "potentially unsafe content".
  • Lost chat history (43 posts, mentioned in 108): chats that vanished, one user reporting about a year of history gone, and history hidden behind the upgrade banner after a redesign.
  • Ads (41 posts): pop-ups that redirect to ad domains, and pages one user says were "smothered" in ads.

What this means for DeepAI Code

DeepAI Code is new, and it would be unfair to blame it for complaints about an image generator. The board has only a few posts about it so far: two users who could not complete sign-in (one of them on Windows), and one asking for a Linux client.

But DeepAI Code shares the same account, the same Pro subscription and the same company, and its frontier models run on a daily Theos allowance with a Pro upgrade on top. The board shows what DeepAI’s users have found frustrating in the past: upgrade prompts, billing they could not easily stop, and features that moved behind the paywall. With a coding agent the stakes are higher, because the thing in the chat is your work.

How Freebuff handles the same things

  • Paying: the Freebuff agent is free with no card. Usage shows as hours, each session is an hour with unlimited messages, and paid plans are there if you want more, not required to code.
  • Quality: Freebuff does one job, coding, and checks its own work. Its code-reviewer subagent critiques changes before you see them, and browser-use tests the app in a real browser.
  • Losing work: your code lives in your own git repository, not in a chat. Freebuff Desktop threads survive a restart, the CLI can save a conversation to a file with /export, and Freebuff Cloud delivers work as pull requests.
  • Where it runs: Linux, macOS and Windows, plus a terminal CLI and the browser. More in our Linux, terminal and browser guide.

One recent post on DeepAI’s board warns the team that maintenance costs are growing faster than ad revenue, then asks if they have heard of Freebuff. We have a full side-by-side in DeepAI Code vs Freebuff.

Try the free coding agent

Terminal, desktop or browser. No card, no API key.

Get Freebuff

Questions

What do people complain about most on DeepAI?
On DeepAI’s public feedback board, the most common complaint is output quality (388 of 1,690 genuine posts), mostly images that ignore the prompt. The most-upvoted posts are about paying: 7 of the top 10 ask DeepAI to make it free or stop the upgrade prompts.
How many posts on DeepAI’s feedback board are spam?
Of the 15,852 posts on the board on November 7, 2026, about 89% were automated spam floods. 1,690 genuine, unique posts remained.
Are these complaints about DeepAI Code?
Mostly not. The board covers all of DeepAI’s products, and most posts are about its image generator, chat and characters. DeepAI Code is new; the few posts about it so far report sign-in failures and ask for a Linux version.
Is DeepAI Pro refundable?
DeepAI’s subscribe dialog says Pro is non-refundable, and billing complaints about charges, refunds and cancelling make up 122 posts on its feedback board.
How did you count the posts?
We collected every post through the board’s public listing, removed automated floods and exact duplicates, then labeled each remaining post with its main theme using AI-assisted labeling that we spot-checked by hand. The counts are close estimates.