ChatGPT Images 2.5 arrived on September 8, 2026, and Google's Nano Banana 2 is already inside the Gemini app, so a US marketer can now get a decent hero image, ad variant or product mock-up without paying anything. The catch is that a free image is not automatically a safe image. Terms differ between tools, some free plans limit commercial use, and the legal status of purely machine-made pictures is unsettled. This guide is about the business side: what to read, what to record and what to review before an AI image appears in an ad, an email or a store page. It reflects public information on September 20, 2026 and is not legal advice.
Start with the licence, not the picture
Every generator publishes terms that decide what you may do with the output. For a US business the questions are short.
- Is commercial use allowed on the free plan? Some tools allow it, while others reserve commercial rights for paid tiers. Reviews in September 2026 note, for example, that Krea's free plan excludes commercial use.
- Who can see your images and prompts? Leonardo AI's free plan is reported to publish images publicly, which is a problem for unreleased campaigns.
- Does the tool claim rights to your output? Read the clause on how the provider may reuse what you generate.
- Are there banned uses? Real people, minors, protected characters and political content are the usual exclusions.
Screenshot the terms page on the day you generate an image. Terms change, and a dated record helps if a client asks.
The copyright question in plain language
US copyright protects human authorship. The US Copyright Office has said that material generated entirely by AI without human creative input generally cannot be registered, while human selection, arrangement or editing may be protected. For a brand this has two practical effects. First, you may not be able to stop a competitor from copying a purely AI-generated logo or mascot. Second, the more you edit, compose and combine, the stronger your position. If a visual is central to your identity, ask a lawyer before you commit to it.
Disclosure and platform rules
Advertising platforms and social networks keep tightening their labelling rules for realistic AI imagery, especially when it shows people or real events. Rules differ by platform and change often, so read the current policy of each channel in your media plan. Both OpenAI and Google now embed provenance signals in their images, C2PA metadata and SynthID watermarks, but these are technical markers, not a substitute for your own disclosure. A simple internal rule works well: if an image could be mistaken for a photograph of a real event or person, label it.
Which tools fit which US use cases
- Ad concepts and social posts: ChatGPT Images 2.5 and Nano Banana 2 handle text in images and multi-step edits well. OpenAI's new Sketch tool lets you draw a layout that the model follows.
- Finished graphics with brand templates: Canva combines a daily allowance of AI generations with an editor.
- Text-heavy graphics: Ideogram is known for legible typography; its free tier reportedly gives 10 slow credits a week.
- Adobe-based teams: Adobe Firefly's free plan reportedly offers 25 credits a month and may watermark output, and Adobe also hosts partner models such as Nano Banana 2.
Limits in free tiers move quickly. Nano Banana 2 free access is capped daily, and reviews quote anything between 20 and 100 images, so verify inside your own account.
A five-step review before publishing
- Rights check. Confirm commercial use on your exact plan and save the terms.
- Content check. Look for real people's likenesses, brand logos, protected characters and stray text.
- Accuracy check. For product visuals, compare the image with the real item. Misleading product images are a consumer-protection risk.
- Disclosure check. Apply the platform label where required and add your own note for realistic scenes.
- Record. Store the prompt, tool, plan, date and final file together.
What it costs when free stops being enough
If you outgrow a free tier, the API route is cheap for batches. Google lists Nano Banana 2 at about $0.067 per 1K image and $0.101 per 2K image on standard pricing, and there is no free API tier for the model. A hundred 1K images would cost roughly $6.70 before retries. Compare that with a stock licence or a photographer's day rate, and account for your review time.
For related reading, see how Google's AI ad formats are changing campaigns, the AI SEO workflow and our guide to AI video with Seedance 2.0, which raises the same rights questions for clips.
An example: a local US service business
Imagine a family-owned landscaping company in Ohio that wants fresh visuals for a spring campaign. The owner uses a free tool for concept images of finished gardens, then hires a photographer for real project photos. In the ads, the AI concepts are labelled as illustrations, while the case studies use genuine photos. The company keeps a shared folder with the terms screenshot, the prompt and the date for each concept. That mix keeps costs low, avoids implying that an illustration is a real customer result and leaves a paper trail if a platform reviewer asks questions.
Mistakes US teams make most often
- Assuming that "free to use" means "free to use commercially".
- Publishing AI images of realistic customers or testimonials, which can mislead audiences and breach ad policies.
- Forgetting that a public gallery may expose an unreleased campaign to competitors.
- Using a generated logo without checking whether it can be protected.
- Skipping a human review of small text, packaging and hands.
- Not saving the prompt and the terms, so nobody can explain later where an image came from.
FAQ
Can a US business use free AI images in ads? Usually yes if the plan's terms allow commercial use, but you must check each tool. Some free plans restrict it or make images public.
Who owns an AI-generated image? It depends on the tool's terms and on how much human creative input went in. Purely machine-made images may not be registrable for copyright in the US.
Do I have to label AI images? Requirements vary by platform and content. Label realistic images of people or events even when a rule is not yet mandatory.
Is a watermark-free free plan safer? Not necessarily. Watermarks and licence terms are separate issues. Read both.
Summary
Free AI image tools are fine for concepts and everyday content, but a US business should treat the licence, the copyright position and disclosure as part of the workflow. Keep a record for every published image and review anything realistic before it goes live. If you want a search plan that connects visuals, content and rankings, see our content optimization and SEO marketing services.
