Shutterstock

Start here, because a great deal of published commentary is wrong about it. The proposed combination of Shutterstock and Getty Images was terminated on 7 July 2026, after Getty declined to accept the editorial-divestment remedy that the UK Competition and Markets Authority required to clear the deal. There is no combined company. Shutterstock trades on its own, sells on its own terms, and competes with Getty exactly as it did before.

That matters commercially, not just journalistically. A merged Shutterstock-Getty would have controlled a large share of Western editorial and commercial stock, and buyers were being told to expect consolidated catalogues and renegotiated enterprise contracts. None of that happened. If you shelved a purchasing decision waiting for the dust to settle, the dust has settled where it started.

Shutterstock launched AI generation in early 2023 with a model trained on its own licensed library — the ethical-sourcing story was the whole pitch. That is not what the product is now.

The current AI image generator is a model selector. Shutterstock's own documentation lists OpenAI's GPT Image 2, Google's Imagen 4 Ultra and Gemini 3.1 Flash, and Runway's Gen-4, plus an AI Model Recommender that picks for you if you would rather not choose. A Prompt Enrichment button rewrites your prompt into something more detailed before it goes to the model.

Read that plainly: when you generate an image on Shutterstock, the inference is happening at OpenAI or Google, and Shutterstock is taking a margin. The output quality is therefore not a Shutterstock property. It is whatever GPT Image 2 or Imagen 4 Ultra does, which you can also get by going to those vendors directly, usually for less money per image.

So the question is not "how good is Shutterstock's model." It is "what does the wrapper add, and is the wrapper worth the markup."

Three things, and they are not trivial.

A licence attached to every download. Shutterstock states that each AI-generated image comes with a Standard License when you download it. That is a document you can hand to a client, a procurement team, or a legal reviewer. Generating the same image at the model vendor leaves you holding an image and a terms-of-service page, which is a materially weaker position when someone asks who owns this.

Indemnity. Shutterstock markets indemnity cover for commercial use of generated content, which is unusual — Adobe is the only other major player that makes this a headline promise. For a company that has been told by counsel not to ship AI imagery without cover, this is the feature that unblocks the project. For a solo creator, it is worth nothing.

One search box over both halves. Generated results and the stock catalogue sit in the same interface, and the editing tools — Magic Brush for inpainting, Expand for outpainting, Variations, Background Remover — work on either. The practical value is the fallback: when the third prompt still hasn't produced a usable product shot, you search the library and license a photograph instead, without changing tabs or vendors. Shutterstock puts the catalogue at 600M+ assets, including 560M+ images, 50M+ videos and 1M+ 3D models.

Pricing is denominated in generation credits, and the arithmetic is worth doing before you subscribe.

As of August 2026 the standalone GenAI plan is $15/month billed yearly — $180 up front — for 100 credits a month. Shutterstock markets that as up to 400 images, on the basis that a generation returns four variations. That is the optimistic reading. The realistic reading is 100 attempts, because iterative work burns a generation per revision and you will rarely keep four outputs from one prompt.

The credits also arrive bundled with the stock plans: 50 a month on Unlimited Images ($29/month billed yearly), 100 a month on Unlimited Plus ($69/month billed yearly). If you already hold a stock subscription, you have generation credits sitting unused right now. Every headline price assumes annual billing; the no-contract monthly rates are higher.

Two free generations are available without a subscription, which is enough to check that the model selector works and not much else.

Shutterstock's library is supplied by contributors — it puts the number at 3M+ across 150+ countries — and its data-licensing arm sells access to that library as AI training data. Named deals include ElevenLabs for sound effects and Lightricks for video training, alongside distribution through Databricks.

Shutterstock's answer to the obvious objection is the Contributor Fund, which pays into the contributor base when library content is used for AI training and data licensing, and a general commitment to "ethically sourced" data. The company does not publish a per-generation royalty rate, and contributors have publicly disputed the adequacy of the payouts since the programme launched. This is not a settled question, and anyone selecting a stock vendor on ethical grounds should read the current contributor terms directly rather than the marketing page.

It is also the sharpest way to state what has changed about the product. The 2023 pitch was that Shutterstock trained on content it had licensed and paid for. The 2026 product routes your prompt to models trained by other companies on data those companies sourced elsewhere. The ethical-sourcing claim now attaches to the data Shutterstock sells, not to the images the tool generates for you.

If image quality is your binding constraint, buy Midjourney and stop reading. If you want the cheapest access to GPT Image 2, go to OpenAI.

Buy Shutterstock if the blocking requirement is paperwork: a licence per asset, indemnity cover, an invoice from a public company, and a procurement process that has already approved the vendor. That describes a lot of in-house marketing teams and most agencies. For them the $15/month GenAI plan is cheap relative to a single hour of legal review, and the stock fallback is worth more than the generator.

Adobe Stock is the direct comparison and the tougher call — similar library, similar generation-plus-stock bundle, stronger indemnification language, and much better integration if the team already lives in Creative Cloud. Shutterstock's edge is the model selector, which lets you route a prompt to whichever frontier model handles it best. Adobe's edge is Photoshop.