Can you use AI-generated content commercially? It is four gates, not one

Search "can AI-generated images be used commercially" and you get an Adobe forum thread, a Gemini support question, a Microsoft Learn post, a Quora answer, a Facebook group, and a Picsart help article. Ten results, two of them actual articles.

That is not because the question is unimportant. It is because the honest answer does not fit in a support-page paragraph. "Commercial use allowed" reads like a yes-or-no property of a tool. It is actually four independent gates, and a tool can let you through the first three and stop you at the fourth.

"Commercial use allowed" answers less than you think

Take Suno, which is unusually clear about this. Its terms of service contain a section headed Commercial Use:

Subject to the Content Section below, unless otherwise expressly authorized herein or in the Service, you agree not to display, distribute, license, perform, publish, reproduce, duplicate, copy, create derivative works from, modify, sell, resell, grant access to, transfer, or otherwise use or exploit any portion of the Service, and any Output or Voice Model, for any commercial purposes.

Read alone, that is a flat no. But it opens with "Subject to the Content Section below," and the Content Section says:

Subject to your compliance with these Terms of Service, if you are a user who has subscribed to the Pro or Premier paid tier of the Service, Suno hereby assigns to you all of its right, title and interest in and to any Output owned by Suno and generated from Submissions made by you through the Service during the term of your paid-tier subscription.

Now look at what that sentence assumes. It does not say the output was yours and Suno is confirming it. It says the output is owned by Suno, and Suno is assigning it to you — but only if you subscribe, and only for what you made while subscribed.

Same product, two customers, opposite answers. Neither is a lie, and a "commercial use: yes" row in a comparison table would be wrong for one of them.

If you want the general method for spotting this kind of structure, we cover it in Who owns what you make with AI?. Here we are staying on the commercial question specifically.

Gate one: free plans are where commercial rights usually disappear

This is the gate people walk into hardest, because the free tier is where everyone evaluates a tool. Suno again, on the free and Basic tiers:

If you are a user of the free or Basic tier of the Service then, you covenant and agree that you will only use Outputs generated from Submissions made by you through the Service solely for your lawful, internal, personal and non-commercial purposes, provided that you give attribution credit to Suno in each case.

Two restrictions in one sentence, and most people only notice the first. Non-commercial use only — and even then, attribution is mandatory.

There is a third trap in the same terms, and it catches paying customers too:

if you are a joint owner of a Remix generated by these Service features, and regardless of whether you are a free Service tier user or a subscriber to a paid Service tier, you additionally covenant and agree that the Remix may only be used for lawful, internal, personal and non-commercial purposes…

So a feature-level restriction overrides your plan. You can be a paying subscriber, believe you have commercial rights, use one particular feature, and land back in non-commercial territory. Plan tier is the usual gate, but it is not the only one — specific features can carry their own terms.

What to check on any tool:

• Does the free tier grant commercial rights at all, or only personal use?
• Is attribution required? Where, and in what form?
• Do rights attach to what you made while subscribed, or to everything in your account? (Suno's assignment is explicitly scoped to "during the term of your paid-tier subscription.")
• If you cancel, what happens to work you made while paying?
• Do any individual features carry restrictions that override your plan?

Gate two: watermarks, resolution and export limits are commercial terms too

These get filed under "features" and priced as upgrades, but functionally they are commercial restrictions. A tool that permits commercial use and exports 720p with a logo in the corner has not given you something you can put in a client deliverable.

Mandatory attribution is the clearest case. Suno's free tier permits use provided you credit Suno. That is not a technical limitation you can work around — it is a contractual condition, and ignoring it is a breach, not a workaround.

Work through the list in the order a real project hits them:

Limit	The question that actually matters
Watermark	Removable on which plan? Is removal permitted, or just technically possible?
Resolution / bitrate	Does the maximum export meet the spec your client or platform requires?
Duration	Per clip or per project? Does stitching clips together change your rights?
Attribution	Required where — in the video, the description, the credits?
Export format	Can you get the file out in something an editor accepts?
Volume	Is there a cap on how much you can produce commercially per month?

A tool passes this gate only if the output you can actually export is the output you can actually ship.

Gate three: nobody warrants that your output does not infringe

This is the gate almost no comparison article mentions, and it is the one with real financial exposure.

The licence answers "may I use this?" It does not answer "will using this get me sued?" Those are different risks, and platforms are explicit about not covering the second. From Suno's terms, in the original capitals:

THE SUNO PARTIES MAKE NO WARRANTY THAT (A) THE SERVICE WILL MEET YOUR REQUIREMENTS; … (D) YOUR USE OF ANY OUTPUT OR VOICE MODEL WILL NOT INFRINGE THE RIGHTS OF ANY THIRD PARTY…

Read that next to the assignment clause. Suno assigns you its rights in the output, and in the same document declines to warrant that using it infringes nobody. Both statements are consistent: a company can only give you what it has. It cannot give you other people's rights, and it is not promising the model never produced something close to somebody else's work.

This is why "trained on licensed data" appears in so much AI marketing — it is an attempt to reduce this specific risk, not the ownership risk. The two get marketed together and they are not the same thing. Ownership is settled by the contract between you and the platform. Infringement risk depends on what the model learned from, and no contract term between you and the platform changes what a court in another country thinks about the output.

What to actually do:

• Look for an indemnity, not a reassurance. Some enterprise tiers offer to defend you if you are sued over output. Marketing copy about training data is not an indemnity.
• If an indemnity exists, read its conditions — they usually require you to have used the tool as directed and not modified prompts to target a specific artist or brand.
• Treat "commercially safe" claims on a landing page as a claim, not a term. If it is not in the terms of service, it is not enforceable.
• For high-stakes work, do a reverse-image or similarity check on the output before it ships.

Gate four: the publishing platform is a separate authority

You have cleared the tool's terms. You still have not cleared the place you are publishing to.

YouTube, TikTok, Instagram, Meta's ad products and Google's ad products each maintain their own rules on synthetic media — disclosure requirements, labelling, restrictions on certain categories (political content and likeness of real people are the common ones), and separate monetisation policies. A tool granting you full commercial rights has no authority over any of them.

Two things make this gate genuinely hard rather than merely annoying:

The rules change often, and they change independently. A disclosure requirement that did not exist when you planned a campaign can exist when you launch it. Anything we wrote here about a specific platform's current policy would have a short shelf life, so we are not going to state one — check the platform you are publishing to, on the day you publish.

Monetisation is a stricter test than permission. Being allowed to post is not the same as being allowed to earn from it. Bulk-produced or minimally-transformed content runs into repetitive-content rules on several platforms regardless of whether the AI part was licensed correctly.

What changes when the work is for a client

Add a fifth party to the chain — the tool, you, the client, and the client's own platform obligations — and two things need to be true that were optional before.

Your rights must be transferable or sublicensable to them. Some tools grant you the right to use output without the right to pass that right down. If your contract promises the client ownership of the deliverable and your tool licence does not permit sublicensing, you have promised something you cannot deliver.

Your warranties must match what you actually have. Standard agency contracts often warrant that deliverables do not infringe third-party rights. You just read the clause where the tool declines to warrant exactly that. Signing the client warranty means you personally absorb a risk the tool explicitly refused. That may be an acceptable business decision — but make it deliberately, and price it.

Practical version: keep a record per project of which tool, which plan tier, which date, and which terms version produced each asset. When a client asks in eighteen months, that record is the only thing that answers them.

A checklist before you publish or sell

Run this once per tool, and again whenever you upgrade, downgrade, or notice the terms page has a new date.

1. Plan tier — do commercial rights attach to my current tier, or only to a higher one?
2. Scope in time — do rights cover work made only while subscribed?
3. Feature-level exceptions — does anything I use carry its own restriction?
4. Attribution — required? Where, and in what wording?
5. Export reality — is the file I can actually produce good enough to ship?
6. Infringement — is there an indemnity, or only a warranty disclaimer?
7. Publishing platform — disclosure, labelling and monetisation rules where this is going.
8. Client chain — can I pass these rights on, and does my warranty match what I hold?
9. Record the date and the terms version. Every answer above expires.

Nine questions, most of them answerable in a few minutes with the search box on a terms page. The alternative is finding out which gate you missed after the campaign ships.

---

A note on sources. All Suno quotations were taken directly from Suno's terms of service as retrieved on 29 July 2026. We did not quote any platform's marketing claims about commercial safety, because marketing pages are not contract terms and the two often disagree.

We attempted to verify Adobe Firefly's public statements about training data directly from Adobe's own documentation and could not retrieve those pages at time of writing, so we have not characterised Adobe's position here rather than relying on second-hand summaries.

Nothing here is legal advice.

Also in this series: Who owns what you make with AI? · fuse Content Policy