Photos and images

Photos are personal data too.

For personal photos, crawler blocking is only one layer. Think about faces, children, location clues, private events, licensing, authentication, watermarks, provenance and platform settings.

Visual assets

Photo protection is layered.

Private photos

Be careful with children, faces, home interiors, events, documents in the background, school uniforms, car plates and location clues. Public images can be copied, so do not rely on one control.

Creator portfolios

Use clear licensing terms, lower-resolution previews, authenticated galleries, watermarks, provenance metadata and platform settings.

How scraping works

How AI image collection actually happens.

Understanding the pipeline helps you see where your own controls can and cannot reach. Much of this is public information about how large image datasets have historically been built, but exact methods differ by company and change over time, so treat this as background rather than a guarantee.

Public datasets and link lists

Many training sets were assembled from large web crawls (for example broad archives of public pages) and from lists of image URLs paired with their surrounding text. If an image sits on a public page that a compliant crawler can reach, it may end up referenced in such a dataset. Removing the image later does not necessarily remove copies already collected.

Alt text and captions

Image models often learn from the words near an image: the alt attribute, captions, file names and nearby paragraphs. Descriptive alt text is good for accessibility and search, but it also makes an image easier to label and index. This is a trade-off, not a reason to remove accessibility text.

EXIF and metadata

Photos can carry EXIF metadata such as camera model, timestamps and sometimes GPS coordinates. Many platforms strip this on upload, but not all do, and files shared directly may keep it. Removing location and identifying metadata before publishing is a sensible habit, though it does not affect what is visible in the image itself.

Reverse image search and re-uploads

Once an image is public, others can download, re-host or reverse-search it. That means the same picture can appear on sites you do not control, each with its own robots rules. This is why no single setting can promise that an image will never be seen again.

Reality check: these controls reduce exposure and signal your intent. They are voluntary conventions and imperfect in practice. Nothing here guarantees an image is excluded from every dataset, and this page is educational information, not legal advice.

Practical checklist

Private photos versus creator portfolios.

For private and family photos

Keep it private: the strongest control is not publishing sensitive photos publicly at all. Use closed albums or trusted sharing instead of open pages.

Mind the background: documents, screens, house numbers, uniforms, badges and car plates can reveal more than the subject.

Strip metadata: remove GPS and identifying EXIF before sharing files, especially images of children.

Check platform settings: review who can see, download or index your uploads, and whether the platform offers any AI or training controls.

For creator portfolios

Show previews, not masters: publish lower-resolution or watermarked versions and keep full-resolution files behind a login or license.

State licensing clearly: a visible usage and licensing statement supports your position, even though it cannot force compliance on its own.

Gate galleries: authenticated or client-only galleries keep work off open, crawlable pages.

Layer your defenses: combine robots controls, provenance metadata and platform settings rather than relying on any single measure.

Provenance and tools

Watermarking, provenance and anti-scrape tools.

Content Credentials (C2PA)

The C2PA standard, surfaced in some tools as Content Credentials, attaches tamper-evident provenance metadata describing how an image was made or edited. It helps others verify origin and can signal authorship, but it does not physically prevent copying, and metadata can be stripped. Support varies by platform, so verify current behavior in your tools.

Visible and invisible watermarks

Visible watermarks deter casual reuse; invisible or forensic watermarks can help you trace copies later. Neither blocks an image from being scraped, and aggressive watermarks affect how your work looks, so weigh the trade-off.

Anti-scrape tools (hedged)

Research tools such as Glaze and Nightshade aim to add subtle perturbations that make images harder for some models to learn from or that disrupt training. Effectiveness varies, is debated, and may change as models evolve. Treat them as experimental extra layers, not guaranteed protection, and read the current project documentation before relying on them.

Platform settings

Where opt-outs and controls usually live.

Many photo and creative platforms have added AI-related settings, but names, defaults and availability change often and differ by region and account type. Use this as an orientation, then confirm the current options in each platform's own privacy or account pages.

Where to lookWhat it may controlCaveat
Account privacy settingsWhether posts and images are public, and who can download or index themPublic content can still be copied by others once seen
AI or third-party data settingsSome platforms offer a toggle to limit use of your content for AI or model trainingAvailability and scope vary; may not cover past collection
Portfolio and licensing optionsRight-click protection, download limits, licensing labelsThese deter casual reuse but are not enforcement
Robots and site controls (own site)Requesting compliant crawlers to skip image directoriesVoluntary; does not remove existing copies elsewhere

Settings and toggles are moved and renamed frequently. Verify the current options in each provider's official help pages before assuming a control exists or covers what you expect.

Related reading

Next steps.

ModelVersus modelversus.com

Compare AI models before choosing a workflow.

ChatGPT Alternatives chatgpt-alternatives.com

Find neutral alternatives when one AI provider is not enough.

AI Subscription Guide aisubscriptionguide.com

Learn which AI subscription fits your real work.

OpenAI crawler docs platform.openai.com/docs/gptbot

Check the current OpenAI crawler and user-agent guidance.

Google crawler docs developers.google.com/search/docs/crawling-indexing/overview-google-crawlers

Verify Google crawler names, indexing behavior and AI-related controls.

Anthropic support support.anthropic.com/

Check current Claude account, data and crawler guidance.

FAQ

Common doubts.

Can I stop AI training on my personal photos?

You can reduce exposure, but you cannot guarantee full control once photos are public. Keep sensitive photos private, review platform settings, avoid public high-resolution uploads and use crawler controls for your own website.

Should I paste my CV or resume into an AI chatbot?

Only after removing details you do not need for the task. Names, addresses, phone numbers, birth dates, signatures, employer secrets and references can often be replaced with placeholders.

Can I upload family documents to AI?

Be careful. Medical letters, passports, school documents, legal papers, bank files and family records can contain sensitive data about other people. Use approved privacy settings or do not upload them.

Do private prompts become AI training data?

It depends on the product, account type and settings. Some providers offer controls, team plans or API terms that treat data differently. Check the current settings before typing sensitive information.

Can I completely stop AI companies from training on my website?

Not with one technical switch. robots.txt can request that compliant crawlers stay away, but it is voluntary and does not remove copies that already exist elsewhere.

Does robots.txt legally block AI training?

robots.txt is a technical convention, not a contract by itself. It can support your policy position, but legal questions depend on jurisdiction, terms, copyright, contracts and enforcement.