On July 14, 2026, Google confirmed two changes to how images surface in search: image generation inside AI Overviews, and a redesigned Google Images home. Neither announcement said original publisher images will be ranked or cited preferentially. What both do is expand the number of surfaces where imagery appears — and make the question "can you prove this image is yours" considerably more practical than philosophical.
That is the shift worth acting on. When generated imagery is effectively free, the images that carry weight are the ones you can document: a real job site, a real team, a real result, with a record of who captured it and under what rights. This is a checklist for producing and documenting those images, and for being honest about what you can measure afterwards.
What the announcement actually supports
Reading Google's own documentation rather than the commentary around it, three things are supportable. Images should be relevant to the page they sit on and genuinely useful to the reader. Technical fundamentals still apply: descriptive filenames, meaningful alt text, appropriate formats and sizes, crawlable URLs, and inclusion in an image sitemap where relevant. And image content is subject to the same quality and spam expectations as text.
What is not supportable is any claim that Google favours original photography, that provenance metadata is a ranking input, or that AI experiences preferentially cite publishers who use in-house imagery. Build the workflow because it makes your content more credible and more defensible, not because someone promised a ranking lift.
Original image formats that earn their place
Not every original image is useful. These formats consistently do work that stock cannot:
- Process documentation. The actual steps of the work, in sequence, on a real job.
- Before and after pairs. Same angle, same framing, dated — the most persuasive image type in service marketing.
- Annotated screenshots. Your own dashboards, configurations or reports with the reading marked up and sensitive data redacted.
- Original diagrams. Frameworks, workflows and decision trees drawn to explain your specific method.
- Team and premises photography. Real people in real spaces, which is what verification-minded readers look for.
- Equipment and materials. What you actually use, which quietly signals competence to informed buyers.
- Data visualisations. Charts built from your own first-party data, with the window and source stated.
The nine-field proof record
For every original image you publish, record these nine fields somewhere durable — a spreadsheet is enough to start:
- Image ID — a stable internal reference that survives file renaming.
- Capture date — the date the image was made, not the date it was published.
- Capture location — site, city, or "studio", at whatever precision is safe to record.
- Creator — the named person or contracted photographer.
- Device or method — camera, phone, screenshot tool, or design software.
- Subject and consent — who or what appears, and whether written consent exists.
- Rights and ownership — who holds copyright and what licence you have.
- Editing performed — cropping, colour correction, redaction, compositing, or generation.
- Approved uses — where the image may appear, and any client-imposed limits.
The record takes about ninety seconds per image at the point of capture and is nearly impossible to reconstruct a year later. Fields six through eight are the ones that prevent actual problems.
Publishing controls
Once an image has a record, publishing it well is mostly discipline:
- Descriptive, human-readable filenames that describe the subject rather than a camera serial.
- Alt text that describes the image's informational content for someone who cannot see it — not a keyword list.
- A visible caption where the image carries evidence, stating what it shows and when.
- Embedded IPTC creator, credit and copyright fields, accepting that some platforms strip them.
- Modern formats and responsive sizes, with explicit width and height to avoid layout shift.
- Lazy loading below the fold, eager loading for the representative image.
- A stable, crawlable image URL on your own domain, included in your sitemap where relevant.
- The representative image referenced in the page's Article metadata and Open Graph tags.
Seven-point QA before publishing
- Does the image add information the text does not already carry?
- Is there a completed proof record, including rights and consent?
- Does the alt text describe content accurately without keyword stuffing?
- Are any client names, addresses, account IDs or personal details redacted?
- Is the file size proportionate, and does the page still pass its layout and speed checks?
- Is the image URL stable, crawlable and self-hosted?
- If any part of the image was generated, is that labelled internally and not presented as evidence?
What you can and cannot measure
Search Console reports image impressions, clicks and average position for Google Images, and that is your one reliable quantitative signal. Our own property, for example, recorded 16 image impressions across 11 queries in the July 25 to August 21, 2026 window — a first-party observation about our site, not an estimate of anyone's search demand.
There is no report for image use inside AI experiences. You can look for your imagery manually and record what you find, but that is observation, not measurement, and it should be labelled as such in client reporting. The honest framing is: track image search performance quantitatively, track AI surface appearances qualitatively, and never blend the two into one number.
Check how your images are currently being handled
Before building a workflow, it helps to see the state of what you already publish — alt text coverage, file weights, crawlability and the entity signals around your pages.
Where this fits
Original imagery with a documented origin is part of the same discipline as everything else in answer-engine work: making verifiable claims that a machine or a sceptical buyer can check. That is the core of our AEO services, and the entity and evidence side runs through GEO services. The technical publishing fundamentals sit within our broader SEO services.
Google has not promised anything to publishers who use their own images. Buyers, however, still notice — and so does every reviewer who checks whether your case study photos are real.
