Can AI see your images? How to get your photos used in AI answers

What are the key facts about images in AI answers?

Do AI answers show images from websites?

Yes, at least in Google's AI Mode and in ChatGPT. For other assistants, the companies' own pages we read do not say.

Google AI Mode. In September 2025 Google said people can ask a question in plain words and "get a range of visual results in AI Mode." It added: "Each image has a link, so you can click out and learn more." Google That link is the part that matters to a business. A good photo on your page can send a visitor to your page.

Searching with a photo. In April 2025 Google added multimodal search to AI Mode. Someone can snap or upload a photo and ask about it. Google says Lens "precisely identifies each object in the image," and Gemini can "understand the entire scene." AI Mode then runs several searches at once about the image and the objects in it. Google The answers to those searches still come from web pages, so your page has to describe what is in your photos for it to match.

ChatGPT. OpenAI's help page on web search says: "If a response includes images, select an image to view its source." OpenAI The same page says that to be eligible for ChatGPT search results, a site should allow OAI-SearchBot. It does not say where the images in answers come from or how they are picked. That part is unconfirmed.

Perplexity. Perplexity's help center describes making new images with AI, but the pages we opened do not explain whether or how answers show images from websites. Perplexity We treat that as unconfirmed too.

Do AI crawlers download and look at my images?

Google's do. For OpenAI, Anthropic and Perplexity it is unconfirmed, because their crawler pages never mention images.

We read each company's official crawler page on 11 October 2026:

• OpenAI lists OAI-SearchBot (search), GPTBot (training), ChatGPT-User (live requests from users) and OAI-AdsBot (ad landing pages). The page talks about "content" and web pages. It says nothing about images. OpenAI
• Anthropic lists ClaudeBot (training), Claude-User (live requests) and Claude-SearchBot (search quality). No mention of images or media. Anthropic
• Perplexity lists PerplexityBot (search results) and Perplexity-User (live requests). No mention of images or media. Perplexity

So nobody outside those companies can say from their docs whether these bots download your photo files, look at them, or only read the page text around them.

Google has a named image crawler. Its user agent token is Googlebot-Image, and Google says rules for it affect Google Images, Discover, Google Video and "all features in Google Search where images, logos, and favicons are presented." Google Google's AI Overviews and AI Mode are part of Search, and Google says they have "no additional technical requirements." Google The crawler page does not name AI Overviews or AI Mode by name, so read the Googlebot-Image scope as Google's wording, not a promise about each AI feature.

Apple says the X-Robots-Tag header is "useful for non-HTML resources (such as PDFs or images)" when you want to give Applebot rules. Apple

How does an AI engine know what is in my photo?

Mostly from the words around it. Google says this directly, and the other AI companies don't say anything.

Google's image guide says it "extracts information about the subject matter of the image from the content of the page, including captions and image titles." Google It calls alt text "the most important attribute" for describing an image. Its example of good alt text is alt="Dalmatian puppy playing fetch", against just alt="puppy", and it warns against stuffing alt text with keywords. Google also asks for "filenames that are short, but descriptive" rather than names like image1.jpg, and for images "placed near relevant text." Google

Text inside images. A photo of your price list, a menu saved as a picture, or a banner with your opening hours puts facts where a text reader can't reach them. None of the AI crawler pages we read says the bot reads text inside pictures. Google's image guide warns against using "an image with text" as the main picture in schema.org markup or the og:image tag. Google

Images that only exist in code. Google says it finds images in the src attribute of an <img> element and that "Google doesn't index CSS images." Google A hero photo set as a CSS background is invisible to Google Images. If your images only appear after JavaScript runs, see Can AI crawlers read JavaScript websites?

How do I get my photos used in AI answers?

Use real photos, put them in normal HTML, and describe each one in plain words on the page. Steps 1 to 8 come from Google's own image guidance. Steps 9 and 10 are our advice.

1. Use a real <img> tag. Google finds images in the src attribute, not in CSS backgrounds. If you use srcset or <picture> for different screen sizes, Google says to always keep a src fallback. Google
2. Write alt text that says what is in the picture. Short and specific: "Technician fitting a heat pump outside a brick house," not "HVAC" or a list of keywords.
3. Name the file in words. heat-pump-install-tampa.jpg beats IMG_0423.jpg.
4. Put the photo next to the text it belongs to, with a caption where it helps. Google reads captions and nearby content to work out what the image is about.
5. Use a common format. Google supports "BMP, GIF, JPEG, PNG, WebP, SVG, and AVIF," and the file extension should match the file type.
6. Choose good, clear photos. Google says "High-quality photos appeal to users more than blurry, unclear images." It also notes images are often the biggest part of page weight, so compress them. Google
7. List important images in a sitemap. An image sitemap helps Google find images it might miss, such as ones loaded by JavaScript. Each page entry can hold up to 1,000 images. Google has removed the old caption, title, geo-location and license tags, so only <image:image> and <image:loc> are needed. Google
8. Mark the main image of the page. Google suggests the schema.org primaryImageOfPage property, which schema.org defines as "the main image on the page." Google schema.org For more on structured data, see Does schema markup help AI search?
9. Use your own photos of your own work. Stock photos appear on hundreds of other sites, so there is little reason for an answer to credit yours. (Our advice.)
10. Don't block the image crawler by accident. Check that robots.txt does not disallow your image folder or Googlebot-Image. If you rely on a CDN or security tool, check it isn't blocking crawlers either. (Our advice.)

Can I keep my images out of AI answers or AI training?

Yes for Google's search features and for Google's and Apple's AI training. For other AI companies, there is no image-only switch in their docs.

• Smaller previews. The max-image-preview robots rule takes none, standard or large. Google says it covers "all forms of search results" and names Google web search, Google Images, Discover and Assistant. It does not name AI Overviews or AI Mode. Google
• Keep a page's images out of Google. The noimageindex robots rule tells Google not to index the images on that page. Google
• Block one image or folder. A robots.txt rule for Googlebot-Image affects every Search feature that shows images. Google
• Gemini training and grounding. The Google-Extended token controls whether your content is used to train future Gemini models and for grounding, which Google describes as giving content from its Search index to the model at prompt time. It has no separate crawler and "does not impact a site's inclusion in Google Search." Google Google's page does not single out images.
• Apple training. Disallowing Applebot-Extended opts your content out of training Apple's foundation models. It does not crawl pages itself and does not remove you from search results. Apple
• OpenAI, Anthropic, Perplexity. Their pages describe robots.txt rules per bot, such as GPTBot or ClaudeBot for training. None offers an image-only setting. OpenAI Anthropic Perplexity
• A rights note inside the file. The IPTC photo metadata standard has a Data Mining field with values such as "Prohibited for AI/ML training." IPTC None of the AI company pages we read says it reads this field. Treat it as a statement of your terms, not a block.

How do I get credit for my photos, and should I label AI-made images?

Add licence details so Google can show who owns a photo, and label AI-made images so people can check them.

Licence details. Google Images can show a Licensable badge that links to your licence terms. To be eligible "you must include the license property" in ImageObject structured data, or fill the IPTC Web Statement of Rights field inside the image file. acquireLicensePage, creditText, creator and copyrightNotice add more detail. If the two sources disagree, Google uses the structured data. Google IPTC describes the Credit Line field as how the image "should be credited when published." IPTC Google's page talks about Google Images. It says nothing about AI answers crediting you.

AI-made images. C2PA is "an open technical standard" for showing where content came from and how it was edited. Its steering committee includes Google, OpenAI, Microsoft, Meta, Amazon and Adobe. C2PA Google said in September 2024 that when an image carries C2PA data, its About this image feature shows whether it was "created or edited with AI tools," in Google Images, Lens and Circle to Search. Google In November 2025 Google added a way to ask the Gemini app whether an image was made by Google AI, by checking for its SynthID watermark. Google IPTC recommends the Digital Source Type value trainedAlgorithmicMedia ("Created using Generative AI") for generated images. IPTC

None of these pages says a label helps or hurts how often an image is shown in AI answers. That part is unconfirmed.

What should be on my image checklist for AI answers?

How do I check whether AI answers use my images?

Ask the questions your customers ask, and look at the images in each answer. In Google's AI Mode, tap an image to see which site it links to. In ChatGPT, select an image to see its source, which OpenAI's help page says you can do. OpenAI Note which answers show a competitor's photo where yours could be, and which of your pages has the best photo for that question. Then check that page against the checklist above. Answers change from day to day, so check the same questions more than once.

Our free AI visibility check fetches your pages the way AI crawlers do, so you can see what they can read without looking at a single picture. Our Prompt Simulator then asks the AI engines your customers' questions and shows which sources each one cites, so you can see whether the fix landed. Videos follow different rules; see How to get your videos cited by AI.

Sources

1. AI Mode can now help you search and explore visually — Google. Read Oct 11, 2026.
2. Bringing multimodal search to AI Mode — Google. Read Oct 11, 2026.
3. Searching the web with ChatGPT — OpenAI. Read Oct 11, 2026.
4. Overview of OpenAI Crawlers — OpenAI. Read Oct 11, 2026.
5. Does Anthropic crawl data from the web, and how can site owners block the crawler? — Anthropic. Read Oct 11, 2026.
6. Perplexity Crawlers — Perplexity. Read Oct 11, 2026.
7. Can Perplexity generate images? — Perplexity. Read Oct 11, 2026.
8. Image SEO best practices — Google Search Central. Read Oct 11, 2026.
9. AI features and your website — Google Search Central. Read Oct 11, 2026.
10. Google's common crawlers — Google Search Central. Read Oct 11, 2026.
11. Robots meta tag, data-nosnippet, and X-Robots-Tag specifications — Google Search Central. Read Oct 11, 2026.
12. Image sitemaps — Google Search Central. Read Oct 11, 2026.
13. Image metadata in Google Images — Google Search Central. Read Oct 11, 2026.
14. primaryImageOfPage — schema.org. Read Oct 11, 2026.
15. About Applebot — Apple. Read Oct 11, 2026.
16. IPTC Photo Metadata User Guide — IPTC. Read Oct 11, 2026.
17. Coalition for Content Provenance and Authenticity — C2PA. Read Oct 11, 2026.
18. How we're increasing transparency for gen AI content with the C2PA — Google. Read Oct 11, 2026.
19. How we're bringing AI image verification to the Gemini app — Google. Read Oct 11, 2026.