The desk

Inclusion policy

Effective 2026-10-22. This is the rule for every tool, including ones the directory owner builds. The same text is stored for machines at content/inclusion-policy.json. Cite this page as writerstoolbelt.com/inclusion-policy.

Not listed

A tool is not listed if any of these is true.

1. (a) There is no real, working product the public can use. An invite-only build or a waitlist does not qualify. The product has to be usable now.
2. (b) AI is not part of the writing workflow. Plottr is the standing example: that app ships without AI, so it is not listed.
3. (c) The public record shows clear deception, including fake reviews, fake company details, or misleading claims about what the AI does, or the terms take ownership of the writer's work.
4. (d) There is no privacy policy at all, or writers' content is used to train models and the company does not offer an opt-out to every user. An opt-out is a setting, or a request route the company opens to everyone. An opt-out that exists only on a paid plan still counts. A legal right to object that is available only to residents of the EU or the UK does not.

Disclosed, not rejected

Everything else is written on the tool page. It is not a reason to leave a working product off the list.

• Availability and regions. Where a tool can be used, including country limits, is stated on the tool page. It is not a reason to leave the tool off the list.
• Platforms. Web, desktop, mobile, or a single operating system is disclosed, not used as a bar to entry.
• Languages. A tool that works in one language, or only in one script, is still eligible. The limit is written on the tool page.
• Launch date. A new product is marked on its page as “New: launched <month year>”. Being new is not a reason to reject it.
• Independent reviews. If none can be found, the page says “No independent reviews found as of <date>”. Absence of outside reviews is disclosed, not punished.
• AI training and retention. The page says what the privacy policy, terms, or FAQ actually say about training and retention. That includes “Policy does not say” and “third-party providers process requests under their own terms”. Silence, or a hand-off to a provider’s own terms, is disclosed. It is a reason to reject the tool only when there is no privacy policy, or writers’ content trains models and the company does not offer an opt-out to every user, as defined in rule (d).
• Affiliate links and conflicts of interest. Affiliate links are labeled. They do not decide which tools are listed, how they are reviewed, or their scores. ProseEngine is the directory owner’s product. It is listed on the same terms as every other tool, and that conflict is disclosed on the site and on its page.

Removed tools

These were listed, then taken off the shelves under the rule named here. The old address stays up and says why.

• Ideogram. Removed 22 October 2026 under rule (d). Its privacy policy (15 January 2025) says prompts and uploads are used to train its models, and no opt-out setting is documented (only a legal right to object by email).
• Midjourney. Removed 22 October 2026 under rule (d). Its EU AI Act training-content summary (March 2026) confirms that user prompts and images train its models, and no opt-out setting is documented (only an EEA legal right to object).
• Speechify. Removed 22 October 2026 under rule (d). Its Studio terms (9 October 2026) say uploaded content may be used to improve its models, with no opt-out setting documented (only an EEA/UK right to object), and its general terms treat uploaded material as non-confidential under a perpetual, irrevocable, commercial license.

The same check

Every tool gets the same check. Each tool page carries a dated Checked line. A new tool is checked again three months after it is added.

The dated line on a tool page is that check. Prices and features still move, and the vendor's site wins an argument with ours. To suggest a tool, start with the submission form. Before you send it, read this page: what blocks a listing, and what we disclose instead.