Direct answer: the 13 product entry pages from study 104 contain 654 img tags, of which 634 have a usable src. By URL extension: SVG 234 (36.9%), PNG 225 (35.5%), JPG 160 (25.2%), GIF 5 and 10 without an extension; WebP and AVIF are 0. 37 images (5.8%) are external, from four pages; alt is non-empty on only 238 (36.4%), missing on 287 (43.9%) and empty on 129 (19.7%).
Product pages are more image-heavy than homepages and article pages: 13 pages average 48.8 images, with P13 alone carrying 270 and P02 109. SVG is 36.9%, suggesting many icons/illustrations; alt missing at 43.9% is the worst of the three page types.
Method: same product pages, static img only
The sample is the 13 product entry pages from study 104 (captured 2026-11-01; P01–P15 minus P04/P14 with no product link). With no network, the cached HTML is parsed for <img>: the extension comes from the URL path, external means a different registrable domain, and alt is split into non-empty/empty/missing; width+height, loading=lazy and srcset are recorded too. Format and host rules match the homepage image-delivery study 97.
| Item | Rule | Boundary |
|---|---|---|
| Image | img with a non-empty, non-data: src | CSS backgrounds excluded |
| External | Registrable domain differs from the product page | Subdomains count as internal |
| Extension | .jpg/.png/.svg and similar at the end of the path | Not the served format |
| alt | Non-empty / empty string / missing | Copy quality not judged |
Result: SVG 37%, PNG 36%, JPG 25%
| Format | Count | Of 634 |
|---|---|---|
| .svg | 234 | 36.9% |
| .png | 225 | 35.5% |
| .jpg / .jpeg | 160 | 25.2% |
| No extension | 10 | 1.6% |
| .gif | 5 | 0.8% |
| .webp / .avif | 0 / 0 | 0% |
The format mix differs from both other page types: article pages are 80.8% PNG and homepages 43.1%, while product pages are only 35.5% PNG and 36.9% SVG. That does not mean the product photos are vector; many SVGs are icons, arrows and certification marks and should be separated from product photography.
alt: 36.4% non-empty, 43.9% missing
| alt state | Count | Of 654 |
|---|---|---|
| Non-empty | 238 | 36.4% |
| Empty string | 129 | 19.7% |
| Missing | 287 | 43.9% |
This is the worst of the three page types: article pages are 76.8% non-empty and 18.5% missing; homepages 38.0% and 29.1%; product pages 36.4% and 43.9%. Product images drive selection, so missing alt leaves both screen-reader users and image search without the information.
Page-by-page (13 pages)
| Page | Images | SVG | PNG | JPG | External | alt non-empty | Missing alt |
|---|---|---|---|---|---|---|---|
| P01 | 4 | 0 | 3 | 1 | 0 | 2 | 2 |
| P02 | 107 | 95 | 6 | 0 | 0 | 29 | 3 |
| P03 | 28 | 0 | 10 | 18 | 24 | 37 | 0 |
| P05 | 16 | 0 | 16 | 0 | 0 | 16 | 0 |
| P06 | 62 | 0 | 61 | 1 | 0 | 28 | 0 |
| P07 | 28 | 0 | 27 | 1 | 3 | 33 | 0 |
| P08 | 17 | 0 | 13 | 4 | 0 | 0 | 1 |
| P09 | 22 | 0 | 21 | 1 | 0 | 2 | 20 |
| P10 | 19 | 0 | 1 | 17 | 0 | 15 | 4 |
| P11 | 35 | 0 | 34 | 1 | 2 | 35 | 0 |
| P12 | 8 | 0 | 0 | 0 | 8 | 1 | 6 |
| P13 | 270 | 139 | 21 | 110 | 0 | 27 | 242 |
| P15 | 18 | 0 | 12 | 6 | 0 | 13 | 9 |
Images concentrate heavily: P13 alone has 270 (42.6% of all) and P02 107; the rest carry 4–62. P13 also has the most missing alt (242); P09 has 20 of 22 missing. P03 has 24 external images and P12's eight are all external, showing that some templates put product images on a third-party CDN.
Compared with homepages and article pages
| Metric | 13 product pages | 14 homepages | 36 article pages |
|---|---|---|---|
| Images with src | 634 | 538 | 499 |
| PNG | 35.5% | 43.1% | 80.8% |
| JPG | 25.2% | 31.4% | 10.8% |
| SVG image files | 36.9% | 22.7% | 0% |
| External | 5.8% | 0% | 37.7% |
| alt non-empty | 36.4% | 38.0% | 76.8% |
| alt missing | 43.9% | 29.1% | 18.5% |
| width+height | 1.1% | 3.0% | 0% |
| lazy / srcset | 0 / 0 | 0 / 0 | 0 / very few |
The three page types fail differently: article pages have too much PNG but decent alt; homepages self-host with low alt; product pages have the most images, the highest SVG share and the worst alt coverage, while rarely self-hosting. One image standard cannot cover all three.
What this means for acceptance
- Check product-image alt separately. 43.9% missing is the worst of the three; start with the main product, detail and certification images.
- Separate icon SVG from product photos. 36.9% SVG is not all vector product art; group by purpose in acceptance.
- Prefer WebP/AVIF for product photos. Both are 0; long product lists can add an on-demand format.
- Add width/height and lazy/srcset. Only 1.1% have dimensions and none has lazy/srcset; long lists download full images on mobile.
- Check external product images. Four pages use them (P03 24, P12 8); a CDN change would affect those product pages.
Limits
- 13-page pilot: the same product entry pages as study 104; not all product pages on the sites and not an industry estimate.
- Extension only: images were not requested, so the served format, bytes and availability are unknown; SVG may be icons or art.
- alt states only: copy quality is not judged and an empty alt may be decorative.
- Single snapshot: cached 2026-11-01; no redesign recheck.
- Page level: navigation, footer and icons included; P13's 270 images include template images.
Data and reproduction
Anonymised data: equipment-product-entry-image-delivery-2026.csv (13 rows × 19 columns: img tags, with src, external, hosts, query params, format counts, three alt states, width/height, lazy, srcset). In-repo copy: docs/research/2026-11-product-entry-image-delivery.csv.
Parser: scripts/homepage_image_probe.py (merges article_image_source_probe.py and article_img_attrs_probe.py); related tests: scripts/test_homepage_image_probe.py (6 tests) and scripts/test_article_image_source_probe.py. Run:
python -m unittest scripts.test_homepage_image_probe
Related: the product entry-page AI-readability pilot, the homepage image-delivery pilot, and the article image-delivery pilot.
Sources and definitions
- Sample: 13 product entry-page caches from study 104 (captured 2026-11-01; P04/P14 have no product link).
- Images: static
imgsrc; extension from the URL path; external by registrable domain; alt in three states. - Measured 2026-11-07 with no network requests; site and product URLs are not published.