Here is the line that catches most people out: a 20 US dollar Perplexity Pro subscription lets you generate an image, but it does not let you publish it. Per Perplexity's own help guidance, images made on the Free, Pro and Max plans are for personal, non-commercial use only. The moment that picture lands on a monetised blog, a client deck or a product listing, you are outside the licence the standard plans grant. That single distinction reshapes the answer to a deceptively simple question.
So yes, Perplexity can create images from a text prompt, the way ChatGPT and Gemini do. But the mechanics, the rights and the limits differ enough from a dedicated image tool that the feature deserves a closer look before you build any workflow on it. This guide covers what Perplexity's image generation actually does in mid-2026, which plans and models support it, how to use it, the three limits that frustrate people most, and the alternatives if images are the point of the work rather than an afterthought to research.
Perplexity does not build its own image model
The most useful thing to grasp first: Perplexity does not train an image generator. It acts as a routing layer that hands your prompt to a third-party model and returns the result inside its answer interface. This mirrors how it handles text, where it orchestrates frontier models from other providers rather than running a foundation model of its own. The practical consequence is that the quality, style and behaviour of a generated image depend entirely on whichever external engine is selected at that moment.
As of mid-2026, the model picker for Pro and Max subscribers has centred on three options. The default is Nano Banana 2, Google's image model released on 26 February 2026 under the formal name Gemini 3.1 Flash Image, which independent benchmarks placed at the top of blind text-to-image leaderboards at launch. Alongside it sit Seedream 4.5 from ByteDance, noted for editing tasks but capped at 2048 by 2048 pixels, and GPT Image 2 from OpenAI as an alternative aesthetic. Earlier additions such as FLUX and DALL-E have rotated through the menu over time. You can set a default under Settings, then Preferences on the web, or let Perplexity choose automatically.
Model versions move fast, so the exact names in your picker may differ from any older tutorial. The constant is the architecture: you are choosing among external models, not a single Perplexity-native engine. If you want the wider context on how this routing approach shapes the platform, our explainer on how AI search works covers the orchestration layer in depth.
A distinction worth keeping straight: generated images versus Getty results
One source of confusion is that Perplexity now handles images in two completely separate ways, and conflating them leads to wrong conclusions about rights. The first is generation, the text-to-image feature this guide is about. The second is image search and display: in October 2025 Perplexity struck a multi-year licensing deal with Getty Images to surface high-quality, properly attributed stock and editorial photography inside answers, with an image credit and a link back to the source.
These are not interchangeable. A licensed Getty photo shown in an answer carries Getty's own licensing terms; it is not yours to reuse freely. A model-generated image is governed by Perplexity's usage policy and the underlying model's terms. When you read that Perplexity has improved its image capabilities, check which one is meant before drawing any conclusion about what you are allowed to do with the picture.
What is and is not supported
Because generation is a secondary feature layered onto a search product, the support matrix is narrower than a purpose-built tool. The table below summarises the position based on Perplexity's published help guidance and plan documentation in 2026. Treat any specific number as indicative: Perplexity adjusts models, quotas and regional rollout regularly, so confirm against your own account.
| Capability | Supported? |
|---|---|
| Text-to-image generation | Yes, on paid plans |
| Choice of image model (Nano Banana 2, Seedream 4.5, GPT Image 2) | Yes, on Pro and Max via Settings |
| Access on the web | Yes |
| Access via the public API | No; the Sonar API is search-grounded text, with no stable image endpoint |
| Commercial use of generated images | No on Free, Pro and Max; reserved for Enterprise tiers |
| Full control over aspect ratio and resolution | Limited; output is often flattened to 1:1 |
| Free tier access | Minimal and subject to regional rollout |
Two rows carry the weight. First, image generation is not exposed as a general public API feature. Perplexity's developer-facing Sonar API delivers search-grounded text through an OpenAI-compatible chat endpoint; there is no documented, stable image-generation endpoint, so you cannot reliably script image output the way you can script search. Second, the commercial-use restriction is the single most overlooked limit, covered in detail below.
Which plans and models support it
Image generation is a paid feature. Perplexity's 2026 lineup runs from a Free tier through Pro at 20 US dollars per month, Max at 200 US dollars per month, and Enterprise tiers priced per seat. According to a 2026 Perplexity pricing breakdown, Enterprise Pro is 40 US dollars per seat per month and Enterprise Max is 325 US dollars per seat per month. That spread matters here because the commercial-use licence only appears at the Enterprise end, a roughly tenfold step up from the Pro plan most individuals start on.
The Free tier is built around limited daily Pro searches and does not provide reliable image generation; where any access has appeared on free accounts it has been minimal and gated by region. The practical entry point is Pro, which unlocks the model selector and a working daily allowance. Max extends those limits and tends to receive newer model options first. Perplexity does not publish a fixed, transparent daily image counter, and community testing in 2026 puts the practical Pro ceiling at roughly ten generations a day before throttling, so heavy users can hit a wall without warning.
Worth noting for context: Perplexity has since extended the same routing pattern to video, rolling out Google Veo-based generation to paid subscribers, with Pro accounts receiving a small monthly credit allowance and Max accounts a larger one. It is the same model: Perplexity wraps an external engine and meters access by tier.
How to generate an image in Perplexity
The flow is straightforward once you are on a supporting plan and using the web app. Image generation is a desktop-web feature first, so start there rather than expecting full parity on every mobile surface.
Want to see this in action?
Check how AI models talk about your brand — free, instant, no signup required.
- Open Perplexity in a web browser and sign in to a Pro or Max account.
- Optionally set your preferred image model under Settings, then Preferences, then the image generation option.
- Type a clear, descriptive prompt. Direct instructions such as "Generate an image of..." work better than vague phrasing.
- Submit the prompt and let Perplexity return its text response first.
- Look for the Generate Image control alongside the response and select it to produce the visual.
- Refine with follow-up prompts or a style option if one is offered.
A useful mental model: Perplexity treats the image as an output attached to an answer, not as the primary result. That is why you usually get a text response first and then trigger the image, rather than the image appearing immediately the way it would in a dedicated generator. It is built for the moment you are mid-research and want a quick illustration, not for a focused design session.
The three limits that catch people out
Most of the frustration people report traces back to three constraints. Understanding them upfront saves a wasted afternoon.
1. Commercial use is restricted to Enterprise. Per Perplexity's help guidance, images generated on the Free, Pro and Max plans are for personal, non-commercial use only, with commercial rights reserved for the Enterprise tiers. As an independent review of the policy spells out, even a monetised channel typically counts as commercial use. So the 20 US dollar Pro plan does not license a generated image for a monetised blog, a YouTube thumbnail, a paid client deliverable, a product listing or sponsored content. If you intend to use output commercially, you need an Enterprise plan or a different tool whose licence permits it. This is also why ownership questions around Perplexity sit on shakier ground than ChatGPT or Claude, whose terms explicitly assign output to the user.
2. Output control is limited. Even where an underlying model supports many aspect ratios and high resolutions, Perplexity's interface has tended to flatten results to a square 1:1, sometimes padding with white space rather than honouring a requested ratio. Seedream 4.5 also caps at 2048 by 2048. If precise dimensions, batch variations, consistent characters or fine editing controls matter, a dedicated tool will frustrate you less.
3. Access is web-first, quota-limited and region-gated. There is no reliable general API access, mobile parity can lag, and daily limits are not shown. When users report that image generation has suddenly stopped, the cause is usually not a bug. Through 2026 thousands hit an error stating the feature is not supported in their region, a deliberate rollout decision tied to local licensing and compliance rather than an outage. Spot-checking a feature like this is exactly why one-off tests mislead; for why continuous monitoring beats manual checks in general, see why spot-checking AI visibility fails.
When Perplexity is the right choice, and when it is not
Perplexity image generation is a convenience feature, and judged on those terms it is good. It shines when you are already inside a Perplexity research session and want a quick illustration, a concept sketch or a personal visual without switching apps. For that occasional, non-commercial use it does the job and saves a context switch.
It is the wrong tool when images are the point of the work. If you need commercial rights, exact aspect ratios, consistent characters across a set, iterative editing, batch output or programmatic access, you will be fighting Perplexity's design at every turn. In those cases a purpose-built generator is the better fit, and the table below frames the trade-off plainly.
| If you need... | Reach for... |
|---|---|
| A quick visual during research, personal use | Perplexity (Pro or Max) |
| Commercial rights at consumer pricing | A dedicated generator with clear licence terms |
| Exact aspect ratios, batch output, editing | Midjourney or a direct model interface |
| Programmatic generation via API | A provider with a documented image endpoint |
Perplexity's real strength is the answer-engine side, where it competes with ChatGPT and Gemini for how brands and products get surfaced in responses. That is a separate question from image generation, and a more strategic one for most businesses; our look at Perplexity versus ChatGPT for brand ranking covers which engine recommends what and why.
Alternatives for image generation
If image creation is your main goal rather than a research add-on, the dedicated options give more control and, in several cases, clearer commercial-use terms at consumer pricing. Match the tool to the licence and the level of control you need.
- ChatGPT integrates image generation into a conversational workflow and is a common default for general use.
- Google Gemini pairs Nano Banana directly with the Gemini app and ecosystem, useful if you already work inside Google tools, and gives you the same engine without Perplexity's wrapper or its square-flattening.
- Midjourney remains a favourite for stylised, high-aesthetic output with fine creative control.
- A direct model interface for Seedream, FLUX or similar gives you the underlying engine with full output settings and the model's own licence terms.
Always confirm the commercial-use terms of whichever service you pick before publishing. Those terms vary by provider and by plan, and they change. The mistake people make with Perplexity, assuming a paid subscription equals a commercial licence, is one to avoid everywhere.
The bottom line
Perplexity can generate images, and for a quick personal visual inside a research session it works well. What it is not is a full image-generation platform. It routes to third-party models such as Nano Banana 2 and Seedream 4.5 rather than building its own, it is a web-first feature with no stable public API, output control is limited and often flattened to a square, and crucially the standard Free, Pro and Max plans grant personal, non-commercial use only. Know those boundaries and you will know exactly when to reach for it and when to use a dedicated tool instead.
For brands, the more important question is rarely whether you can make an image in Perplexity. It is how Perplexity and other answer engines describe and recommend your business. That is a monitoring problem, and it is what Honeyb tracks across ChatGPT, Perplexity, Gemini, Google AI Mode and more. Start with a free scan using our AI visibility checker, or see how the leading engines stack up in our best AI search engines guide.
