If the ChatGPT image generator is not working, first separate the problem into one of four buckets: an OpenAI outage, an account or usage limit, a browser or app problem, or a blocked prompt. The right fix depends on which bucket you are in. A generation that spins for a minute is not the same as a missing Create image tool, a blocked prompt, a blank web page, or an API rate-limit error. This guide walks through the practical checks in the right order so you do not waste time clearing cookies when the service is down, or waiting for an outage when your VPN, extension, prompt, or mobile app is the real cause.
Fast diagnosis: outage, limit, local bug, or prompt block
Start with the symptom. “ChatGPT image generator not working” can mean several different failures. The fastest path is to identify the failure type before changing settings.
OpenAI’s help article says ChatGPT image generation can take up to two minutes depending on prompt complexity, so a short wait by itself is not always a failure.[1] OpenAI also documents general ChatGPT failures caused by browser extensions, VPNs, proxies, secure DNS tools, site data, long chats, and transient server-side problems.[3]
| What you see | Most likely cause | First fix to try |
|---|---|---|
| Image tool is missing | Account, rollout, plan, app version, or workspace setting | Open a new chat on the web app, then check the tools menu and your account state. |
| Spinner stays on “generating” | Slow generation, overloaded service, long chat, or local connection issue | Wait briefly, then stop and retry in a new chat. |
| “Something went wrong” or blank page | Browser data, extension, VPN, network, or OpenAI incident | Check OpenAI status, then test in a private window with extensions off. |
| Prompt is refused or blocked | Safety policy or security filter | Rewrite the request in safer, more specific terms. |
| Only mobile app fails | App cache, stale app version, device state, or login session | Try the same prompt in a mobile browser or desktop browser. |
| API returns rate-limit or quota errors | API usage tier, model limit, billing, or endpoint configuration | Check the platform limits page and request logs. |
Do not assume the cause is your account. OpenAI has had feature-specific image generation incidents before. For example, an official status incident on August 4, 2025, investigated ChatGPT image generation errors at 11:56 AM and marked the partial outage resolved at 12:57 PM.[4] For broader downtime patterns, keep our ChatGPT outages 2026 tracker and ChatGPT outage history handy.
Check whether image generation is available in your session
Before troubleshooting the browser, confirm that you are actually in a ChatGPT session that can create images. OpenAI’s image creation help page says you can generate images by asking ChatGPT directly or by selecting the image creation option from the tools menu.[1] If neither approach works, the problem may be session availability rather than a broken prompt.
Use this quick check:
- Open a new ChatGPT chat.
- Type a simple request: “Create a square image of a red mug on a white table.”
- Do not attach a reference image yet.
- Check whether ChatGPT starts image generation, asks a clarifying question, refuses the request, or answers with text only.
- Open the tools menu and look for the image creation option.
If a plain image prompt works, your account and session can generate images. The issue is likely the original prompt, attachment, conversation context, or edit request. If the plain prompt fails too, continue with the device and status checks below.
OpenAI introduced GPT-4o image generation in ChatGPT on March 25, 2025, and said it was rolling out as the default image generator for Plus, Pro, Team, and Free users, with Enterprise and Edu access coming later.[2] That history matters because some older help threads refer to DALL·E behavior, while current ChatGPT image behavior may depend on the newer built-in image flow. If you are comparing your experience with an old screenshot, confirm that you are using the current ChatGPT interface.
Fixes for the ChatGPT web app
The web app is the best place to isolate image generation problems because it removes mobile app variables. Use these steps in order.
1. Check OpenAI status first
If image generation fails in multiple browsers or on multiple devices, check the OpenAI status page before making local changes. OpenAI’s troubleshooting guide tells users to check status.openai.com for ongoing incidents when ChatGPT is slow, frozen, or unresponsive.[3] If there is an active image generation or ChatGPT incident, wait for recovery rather than repeatedly retrying large image prompts.
2. Start a clean chat
Long conversations can create extra context and UI state. OpenAI recommends starting a new chat when a conversation is long or has many turns.[3] For image generation, this is especially useful if the chat contains many attachments, edits, or conflicting visual instructions.
3. Test a private window
Open ChatGPT in an incognito or private window. Sign in. Try the same simple image prompt. If it works there, the problem is probably browser state, an extension, or cached site data. OpenAI’s troubleshooting guide specifically recommends private browsing, different browsers, disabled extensions, and clearing site data for frozen pages or endless loading.[3]
4. Disable extensions and content blockers
Disable ad blockers, privacy extensions, script blockers, grammar tools, download managers, and security extensions for the test. These tools can interfere with web app scripts, uploads, authentication, or streaming connections. If the image generator starts working, re-enable extensions one at a time until you find the conflict.
5. Turn off VPNs, proxies, and secure DNS filters
VPNs and filtering tools can break sessions, trigger additional checks, or block supporting domains. OpenAI’s troubleshooting steps include disabling VPNs, proxies, secure DNS services, and other security filters when ChatGPT is frozen or stuck loading.[3] If this fixes image generation, see our ChatGPT not working with VPN guide for a cleaner long-term setup.
6. Clear site data after testing
If private browsing works but your normal browser does not, clear site data or cookies for the ChatGPT domain, then sign in again. This is more targeted than wiping your whole browser. If login breaks during that process, use our ChatGPT login failed guide.
Fixes for iOS and Android
If image generation works on desktop but not in the app, treat it as a mobile app problem. The model and account may be fine.
- Force close the ChatGPT app and reopen it.
- Switch between Wi-Fi and mobile data.
- Sign out and sign back in.
- Update the app from the official app store.
- Restart the phone.
- Uninstall and reinstall the app if the problem continues.
- Try the same prompt in a mobile browser.
OpenAI’s iOS troubleshooting page recommends relaunching the app, uninstalling and re-downloading it for crashes, and collecting details such as screenshots, date and time, device model, language, custom keyboards, and accessibility settings when the problem persists.[8] Those details are useful because image generation failures can look like model failures when the real issue is app state, upload handling, or a device-specific crash.
If the app cannot load chats, freezes, or crashes outside image generation, use our ChatGPT app not working troubleshooting guide. If files or reference images fail before the prompt even reaches ChatGPT, see ChatGPT file upload not working.
When the prompt or uploaded image is the problem
A blocked prompt is not the same thing as a broken image generator. OpenAI says prompts may be blocked for content violations or security reasons, including requests that could lead to outputs that violate its policies or attempts to manipulate the system.[5] OpenAI also says it uses automated tools to detect prompts, completions, and uploads that may be harmful or violate usage policies, and may warn users or block the model from responding.[6]
Rewrite the request instead of repeatedly resubmitting it. A good rewrite removes risky ambiguity and states a normal creative goal. For example:
| Weak request | Better request |
|---|---|
| “Make this person look worse.” | “Create a fictional character portrait with tired lighting, messy hair, and a dramatic editorial style.” |
| “Copy this logo exactly.” | “Create an original badge with a circular layout, geometric shapes, and a blue-and-gold palette.” |
| “Make a realistic ID card.” | “Create a fictional event pass for a pretend design conference, clearly marked as a sample.” |
| “Turn this celebrity photo into something explicit.” | “Create a fictional adult character in a non-explicit fashion editorial pose.” |
Uploaded images can also cause failures. If you attach a copyrighted asset, a real person, a low-quality screenshot, or an image with confusing visual elements, the system may refuse, misread the instruction, or produce a poor result. Test the same edit with a simpler image. Then add constraints one at a time.
For better image prompts, use concrete nouns, composition, style, aspect ratio, colors, and exclusions. Avoid asking for “better,” “cooler,” or “more viral” with no visual direction. OpenAI’s help page says ChatGPT Images can follow precise instructions to add text, add details, or make backgrounds transparent, so precise instructions are more useful than vague adjectives.[1]
Limits, slow generations, and stuck spinners
Image generation is heavier than ordinary text chat. A delay does not always mean failure. OpenAI says ChatGPT Images may take up to two minutes depending on instruction complexity.[1] The original GPT-4o image generation release also warned that detailed images take longer to render.[2]
If you see an in-product usage limit message, believe the message. OpenAI has not published one universal official ChatGPT image quota that applies to every plan, region, workspace, and account state. The practical fix is to wait until the displayed reset time, reduce rapid retries, or use a plan or workspace with higher limits if your work depends on frequent image generation. For broader account limit context, see our ChatGPT daily limit and ChatGPT Error 429 Too Many Requests guides.
If the spinner appears stuck, use this order:
- Wait long enough to rule out a normal slow generation.
- Stop generation if the UI allows it.
- Click regenerate once.
- Start a new chat and send a shorter prompt.
- Remove reference images and test text-only generation.
- Try a private browser window.
- Check OpenAI status.
OpenAI’s general stuck-response guidance recommends waiting 30–60 seconds, stopping and regenerating, starting a new chat, hard refreshing, signing out and back in, testing private browsing, disabling VPNs or proxies, and trying another browser, network, or device.[3] If the issue looks like a general slowdown rather than an image-only failure, use our why is my ChatGPT so slow guide. If the page returns a server error, compare it with ChatGPT internal server error and ChatGPT network error.
API users: separate checks for image endpoints
If you are using the OpenAI API, do not troubleshoot it like the ChatGPT app. ChatGPT subscription access and API access are different surfaces. A prompt may work in ChatGPT while your API request fails because of endpoint, model, organization, billing, file format, or rate-limit settings.
OpenAI’s image generation rate-limit help page says API image generation limits depend on the model and usage tier, and that developers can check account-specific limits on the platform limits page.[7] That means a generic web article cannot tell you your exact API image quota. Use your account dashboard and response headers or error body.
For API failures, check these items:
- Confirm the request is going to the image generation endpoint, not a chat-only endpoint.
- Confirm the model name is valid for image generation in your account.
- Check whether the account has billing enabled and available quota.
- Inspect the exact HTTP status code and error message.
- Reduce image size, prompt complexity, and batch size for a test request.
- Try a minimal prompt with no uploaded file.
- Check the platform limits page for your organization.
If your code retries instantly after rate-limit errors, fix the retry behavior before raising limits. Aggressive retries can turn a temporary limit into a longer outage for your own app. For cost planning, use our OpenAI API pricing reference rather than guessing from ChatGPT plan pricing.
What to send OpenAI support
Contact support only after you can reproduce the issue. A clear report gets a better answer than “images do not work.” OpenAI’s troubleshooting guide asks users to collect a HAR file, browser console errors, timestamps, the model used, and the conversation URL or ID when problems persist across browsers, devices, and networks.[3]
Send this support packet:
- Your account email.
- Device, operating system, browser, and app version.
- Whether it fails on web, iOS, Android, or all platforms.
- The exact prompt that failed.
- Whether a simple test prompt works.
- Screenshot or screen recording of the error.
- Date, time, and time zone.
- Conversation URL or ID if available.
- Any request ID or error code shown in the interface.
- Whether you use a VPN, proxy, secure DNS, browser extension, or corporate network.
Do not include private documents, sensitive photos, passwords, or payment details in a support message unless OpenAI specifically requests a secure upload path. If the problem is a region or account-access issue, our ChatGPT region not supported guide may help you identify the boundary before you contact support.
Frequently asked questions
Why is ChatGPT answering with text instead of making an image?
You may be in a session where the image tool is unavailable, the model did not infer that you wanted an image, or the prompt was blocked before tool use. Start a new chat and explicitly say “create an image.” If that works, the original conversation context caused the problem.
How long should I wait before assuming image generation is stuck?
OpenAI says ChatGPT image generation may take up to two minutes depending on prompt complexity.[1] If it runs far beyond that, stop the generation, retry once, and then test a shorter prompt in a new chat. If multiple simple prompts fail, check status and local browser issues.
Does clearing cookies fix the ChatGPT image generator?
It can, but it should not be the first step unless private browsing works. Test in an incognito window first. If image generation works there, clear site data for ChatGPT and disable extensions before changing account settings.
Why did ChatGPT block my image prompt?
OpenAI says prompts can be blocked for policy or security reasons, including requests that could lead to harmful, inappropriate, illegal, or policy-violating outputs.[5] Rewrite the prompt around a safe fictional or original creative goal. Do not try to bypass the filter.
Is this the same as a ChatGPT outage?
Not always. Image generation can fail because of a local browser issue, an app bug, a usage limit, a blocked prompt, or a feature-specific OpenAI incident. If it fails across devices and accounts, check the OpenAI status page and our outage trackers.
Why does the image generator work on desktop but not on my phone?
That usually points to the mobile app, device network, app session, or app version. Try the same prompt in a mobile browser. If the browser works, reinstall or update the app and capture screenshots if you need support.