Prospective clients and their families size up a behavioral health website in the first few seconds, and most of that judgment is visual. Before they read a word about levels of care or insurance, they are looking for evidence that a real place exists, staffed by real people. That makes photography one of the highest-leverage assets on a treatment provider’s site — and one of the most commonly mishandled, because images sit at the intersection of three disciplines that rarely talk to each other: privacy compliance, accessibility, and search.
Below is the order we work through those decisions on a healthcare site, because getting the sequence wrong is what creates the expensive problems.
Privacy decisions come before optimization decisions
No amount of image optimization is worth a privacy violation, so this is the first gate, not the last. Under the HIPAA Privacy Rule, a covered entity generally needs a valid written authorization from the individual before using protected health information for marketing purposes. Photographs matter here specifically: the Privacy Rule’s de-identification safe harbor lists full-face photographic images and any comparable images among the identifiers that must be removed for data to count as de-identified. A recognizable client photo published on a program page is, in practical terms, a disclosure that this person received care at that program.
For substance use disorder programs the bar is higher again. The federal confidentiality regulations for SUD patient records at 42 CFR Part 2 restrict disclosures that would identify someone as having a substance use disorder, and the consent requirements there are separate from HIPAA’s. State law can add further restrictions. None of this is legal advice, and the compliance question belongs with the provider’s own counsel — but the marketing consequence is simple enough to act on: build the site’s photo library out of staff, clinicians, buildings, grounds, and empty group rooms, photographed with no clients present.
Two habits make that sustainable. Collect a signed model release for every staff member who appears in a photo, and keep a record of which images each release covers so you know what to pull when someone leaves. And shoot rooms deliberately empty. A well-lit, tidy, unoccupied group room reads as calm and credible; a staged one with actors reads as stock photography, which is the impression you were trying to avoid.
Alt text is a text alternative, not a keyword slot
Alt text exists so that someone who cannot see the image still gets its meaning. The W3C’s WCAG 2.2 Success Criterion 1.1.1 requires that non-text content have a text alternative serving the equivalent purpose, and the useful word in that sentence is purpose. A screen reader user navigating your admissions page does not need a list of nouns. They need to know what the image is doing there.
That leads to three concrete rules. Describe the image as you would to someone on the phone: “Two clinicians talking in a sunlit group therapy room” beats “drug rehab Los Angeles group therapy.” Give purely decorative images — divider graphics, background textures, icons that merely repeat adjacent text — an empty alt attribute, so assistive technology skips them instead of narrating clutter. And when an image is wrapped in a link, the alt text has to describe the link’s destination rather than the picture, because that is the only label the user gets.
Stuffing keywords into alt attributes is the oldest bad habit in image SEO. It degrades the experience for exactly the people the attribute was written for, and it is trivially detectable. Write for the listener and the search benefit follows.
File names, captions, and the text around the image
Google’s guidance on how images appear in Google Search is explicit that page context matters: titles, captions, and the text surrounding an image all help Search understand its subject. Three things follow.
Rename files before upload. A camera produces IMG_4471.JPG; a descriptive file name such as clinical-director-portrait.jpg is free signal, and it makes the media library searchable for whoever manages the site next year. Second, use real captions wherever a caption adds something — the name and credential under a clinician’s headshot, the location under a facility shot. Captions get read far more often than body copy. Third, place images next to the text they illustrate. An image dropped three sections away from the topic it depicts loses the context that gave it meaning.
Compress before you upload, not after
The most common performance problem we find on provider sites is not a slow host. It is a 4,000-pixel-wide, six-megabyte JPEG straight off a photographer’s card, displayed in a container 800 pixels wide. The browser downloads every byte and then discards most of them.
Resize to roughly the largest dimension the layout will ever display, then export to a modern format. WebP and AVIF typically produce substantially smaller files than JPEG at equivalent visual quality, and both are supported across current browsers. As a working target, a full-width hero image should land well under a few hundred kilobytes and an inline content image considerably less. This is ordinary web design hygiene, but it has to be enforced at upload time — once a heavy library exists, cleaning it up becomes a project of its own.
Do not lazy-load the image at the top of the page
Lazy loading defers offscreen images until the visitor scrolls toward them, which is the right default for a long page. Applied to the hero image it is actively harmful, because the browser then has to run layout before it discovers the single largest thing it needs to paint. That directly delays Largest Contentful Paint, the Core Web Vitals metric most often decided by a hero image.
The fix is a split: mark below-the-fold images with loading=”lazy” and leave the hero eagerly loaded, optionally with fetchpriority=”high” so the browser fetches it ahead of less critical resources. If the site lazy-loads through JavaScript rather than the native attribute, Google’s documentation on lazy-loading content correctly is worth reading closely — an implementation that only fires on a real scroll event can leave images invisible to crawlers entirely.
Serve every device a size it can use
Responsive images let a single img element offer several files and let the browser choose, using the srcset attribute to list the candidates and sizes to describe how wide the image will actually render. WordPress generates these variants automatically on upload, so most provider sites already have the machinery installed. What breaks it is a theme or page builder that hardcodes the full-size URL, leaving a phone on a cellular connection to download the desktop file. Check a rendered page’s markup directly rather than assuming the platform handled it.
Give machines the structured version
Two additions make images legible to systems that never render your page. Structured data can carry image references explicitly — Schema.org’s ImageObject type supports properties including caption, contentUrl, creditText, and license, which is how you attribute a photographer or state usage terms in a machine-readable way. And an image sitemap, or image entries within your existing sitemap, hands Google a direct list of image URLs it might otherwise miss on a script-heavy gallery.
This is also where images meet answer engine optimization. AI search surfaces answers as text, so the parts of an image that can travel into a generated summary are the parts written in words: the alt text, the caption, the heading above it, the credential printed beside the headshot. A photograph of your clinical director conveys nothing to a language model. A caption reading “Board-certified in addiction psychiatry, medical director since 2019” conveys the whole point of the photograph.
A pre-publish checklist
- No identifiable clients or patients in any image, in any format.
- Signed model release on file for every staff photo, with a record of which images it covers.
- Descriptive file name, set before upload.
- Alt text that states the image’s purpose; empty alt on decorative images; destination described when the image is a link.
- Resized to display dimensions and exported to WebP or AVIF.
- Hero image eagerly loaded; everything below the fold lazy-loaded.
- srcset present in the rendered markup, verified on the live page.
- Captions carrying names, credentials, and locations wherever they exist.
None of this is glamorous, and none of it changes visibility overnight. What it does is remove a category of quiet failures — the privacy exposure nobody reviewed, the hero image adding a second to every page load, the eight staff photos labeled IMG_4471 — that otherwise sits on a site for years. If you want a second set of eyes on how your image library is affecting performance and visibility, that review is part of what an SEO engagement covers.
