Why this signal matters for Google
A stock photo proves one thing: that you paid for a subscription. The handshake on a white background, the team laughing at a whiteboard, the tradesperson in a spotless hard hat: those images run on hundreds of sites in the same trade. Your visitor already saw them at your three competitors. A photo you took says something else: this is my workshop, this is Tuesday's job, this is the machine with the scratch on the lid. It works the same way as a lived testimonial. What counts is not that it looks good, it is that nobody else could publish it.
Google has humans check that difference. Evaluators rate page quality using a public manual, the Quality Rater Guidelines (QRG for short). Section 3.4 asks them to look for real experience with the topic, not knowledge copied from elsewhere. Section 7.3 recognises everyday experience: the kind that belongs to someone who has actually done the thing, degree or no degree. A photo taken on site is the fastest version of that proof to read. The evaluator does not have to take your word for it, they can see your premises, your tools and your work in progress.
The check is mechanical, and you can run it yourself. Two tools are enough. Reverse image search first: drop your image into Google Images and you see every site showing the same one. If it turns up on forty other pages, it is stock, and anyone finds that out in ten seconds. Then EXIF: the information your camera writes into the file at the moment of the shot, date, time, model, sometimes the location. A stock image has empty EXIF or fields filled in by a studio. One warning: many publishing tools strip that data on upload, so check the image once it is live, not on your own drive. The same reflex applies to your product photos, where the manufacturer's supplied visual is the strongest temptation.
On sensitive topics, a generic image costs more. If your site deals with health, money, law or safety (Google calls those topics YMYL, for Your Money or Your Life), section 3.4.1 expects real trustworthiness. A clinic that fronts its homepage with a smiling doctor bought online sends the wrong message: you see no practice, no team, no equipment. On the AI side the problem is different but it points the same way. When ChatGPT or Google's AI summaries pick who to quote, they look for what exists in one place only. An image shared with two hundred sites adds nothing. Your photos, your captions and your alt text (the short description of the image written into the code) do.
The NeuroEEAT scoring grid (0-10)
Rate what your images actually prove to a stranger who spends ten seconds checking.
| Score | Description | Concrete indicators |
|---|---|---|
| 0-2 | All bought | Every image comes from a stock library or a generator. None shows your premises, your products or your work. A reverse image search finds all of them elsewhere. |
| 3-4 | One or two real photos | A few in-house shots, usually pushed to the bottom of the page and drowned in bought illustrations. Nothing says when or where they were taken. |
| 5-6 | Real photos, no markers | Most images are yours, but with no caption, no date, no place, empty alt text and EXIF stripped on upload. |
| 7-8 | Documented photos | In-house photos on the pages that matter, captions with date and place, descriptive alt text, EXIF preserved, and no image turning up elsewhere in a reverse search. |
| 9-10 | Full proof | Every important page has its own images, taken on different dates, tied to the author who signs the page and consistent with the rest of the content: before and after, logbook, video. Files are named legibly and nothing repeats from one page to the next. |
How to audit it on your site
Five checks to find out what your images really say. Start with the ten pages that get the most visits.
- 01
Run your images through reverse search
Open Google Images, click the camera icon, drop in the main image of each page. If the same image shows up on other sites, it proves nothing. Write down how many pages are affected: that is your starting point.
- 02
Open the EXIF of an image already online
Download the image from your public site, not from your computer, then look at its properties or run it through ExifTool. If everything is empty, your publishing tool strips the data on upload and your real photos look exactly like stock.
- 03
Count what each page actually shows
For every page, separate the images that show your reality (your premises, your hands, your results) from the ones that decorate. A service page without a single image of your own work is a page that talks without showing.
- 04
Read your file names and your alt text
A file called IMG_4472.jpg with empty alt text tells Google nothing, and tells a blind visitor using a screen reader even less. List every image concerned: it is the fastest fix on this list.
- 05
Check the photo matches the text
The photo has to show what the page is about. A real photo off topic is no better than a bought one: if your page covers floor laying, show floor laying, not the team portrait already used three times.
How to activate or improve it
Three levels, from your first batch of phone photos to a habit you keep without thinking about it.
Take your first real photos
- Walk through your business for an hour with your phone: premises, tools, products, one task in progress.
- Shoot near a window, no flash, and take three framings of each scene so you have a choice later.
- Replace the bought images on your three most visited pages first, not on the blog.
- Write a caption with the date and the place, along the lines of "Leeds workshop, March 2026".
- Rename each file before upload: oak-floor-laying-living-room.jpg rather than IMG_4472.jpg.
Make the photo verifiable
- Write alt text that describes the scene in one sentence, not a list of keywords.
- Check the EXIF after publishing and change your export settings if your tool wipes them.
- Take one photo at the start and one at the end of every job: you get a before and after for free.
- Date your batches and spread them out: ten photos shot in one day covering twenty pages is visible.
- Sign visually without ruining the image: a caption naming the person on site beats a watermark across half the frame.
Turn it into a system
- Build photography into the job itself: one mandatory shot per visit, dropped into a shared folder.
- Keep an internal library sorted by page, so the same image never covers two different subjects.
- Tie each photo to the author who signs the page, so the visual proof and the identity proof tell one story.
- Add images to your markup (ImageObject in the JSON-LD code) only when they are genuinely displayed on the page.
- Re-run your pages through reverse search once a year: bought images always creep back in during redesigns.
5 common mistakes to avoid
Sector specifics
Practitioner or clinic
A topic with real risk for the reader, where section 3.4.1 expects real trustworthiness. Show the place, the waiting room, the equipment, the team on reception. Never show an identifiable patient without written consent, and never stage a procedure you do not perform.
This is where it is decided
The customer nearby wants to see the shopfront, the inside and the finished work. A photo taken on site, with a recognisable street, also helps your visibility in nearby searches. The same images should appear on your business profile and on your site.
Escaping the supplier catalogue
Manufacturer visuals are identical at every reseller, so they do not set you apart. Add at least one photo of the product at your place, unboxed, held in hand, next to an everyday object for scale. On an otherwise ordinary product page, that is often what makes the difference.
Showing invisible work
A firm or an agency has no product to photograph, but it has screens, boards, workshops and client meetings. An anonymised screenshot of your real dashboard proves more than a bought office scene.
Signals to activate in parallel
Original photos are worth most when they lean on the other experience signals in the table:
FAQ
Original photos vs stock photos: can Google really tell them apart?
Yes, and it is not alone. A stock image is indexed across hundreds of sites, which makes it instantly recognisable through a reverse image search. Section 3.4 of the QRG asks evaluators to look for traces of real experience, and a photo taken on site is the fastest one to read. Your visitor, meanwhile, has often seen it somewhere else already.
My phone photos look worse than the stock ones. Does that hurt?
No. A slightly rough photo of your actual workshop convinces more than a perfect visual seen everywhere. Three rules cover most of it: daylight, a sharp subject, a tidy frame. Keep the polished look for the homepage and the honest look for the pages that explain your work.
Do AI generated images count as original photos?
No. They are unique, but they prove no experience: nobody was on site. You can use them as decorative illustration if you say so plainly. Passing them off as real photos is the most expensive risk on this list, because one image caught out casts doubt on everything else on the page.
Is EXIF data essential?
It helps, but it is neither sufficient nor sturdy. Many publishing tools strip it automatically, and it can be edited. Treat it as one more clue, not as proof. A dated caption, descriptive alt text and the fact that the image exists nowhere else carry more weight.
How many original photos does a page need?
One real photo in the right place beats six decorative images. On a service page or a guide, aim for two or three images that each show a different stage of the work. Beyond that you slow the page down without proving anything more.
Anonymized case study
Before · an interchangeable site
After · one afternoon of shooting and two hours of uploading
Illustrative figures meant to show the typical progression of the signal. They do not come from a real client case and must be replaced with your own measurements before any external communication.