Research Guides
Reconcile the Image Proposed to Represent a Research Article
A research article can show a relevant illustration while its preview metadata still points to an old campaign banner. Checking the visible page alone misses the mismatch. Review which image the publisher is proposing to represent the article, and keep that proposal separate from what a search engine actually displays.
This worksheet concerns preferred preview imagery. It is not a promise of a particular search result, a ranking improvement or automatic image generation by Colquhoun.AI.
1. Separate preference from selection
Google's image SEO documentation says preview selection is automated and considers multiple sources. Publishers can express a preferred image through supported structured data or an og:image meta tag. Google recommends a relevant, representative image rather than a generic site logo or text-heavy image, avoiding extreme proportions and using high resolution when possible.
That guidance provides a review target, not a display contract. A valid image URL in metadata establishes what the page declares. It does not establish that Google has fetched the latest page, selected that asset or shown it for every query.
2. Audit a fictional reused article template
Imagine a fictional website research article explaining how a source-review process works. Its body contains an original illustration of the process. The template's og:image still references a promotional graphic for an unrelated seasonal sale, while its article structured data references a third image.
The reviewer records all three assets and the role each is playing. Instead of changing every image field to the site's logo, the editor chooses a representative asset for this article and reconciles the relevant declarations. The illustration still needs its own rights and accessibility checks; metadata consistency does not replace them.
The repair ticket should identify the incorrect template value and the intended article-specific choice. That makes the fix reproducible when the same template publishes its next article, rather than solving only one URL by hand.
3. Build a preferred-image record
- Article: canonical page address and the subject it actually covers.
- Declared candidates: image URLs found in relevant structured data and preview metadata.
- Visible context: whether each candidate represents the article or unrelated promotion.
- Asset check: final URL, response status, format and dimensions.
- Ownership: who can approve the image and verify reuse rights.
- Decision: selected representative asset and the fields to reconcile.
Record redirects rather than assuming a familiar filename still identifies the same graphic. Keep this inventory distinct from responsive-image testing: the latter checks the versions delivered within the page, while this review checks the image proposed to represent the page elsewhere.
4. Verify the declaration and the actual asset
After the change, retrieve the public article and inspect the metadata actually served. Then open the intended public image through an appropriate safe viewing method and confirm that it is the approved asset, not an error page or a stale campaign graphic.
Record both the technical result and the visual inspection. A successful image response does not tell a reviewer whether a crop removes the subject or makes the picture misleading. If only headers were checked, describe that limited check accurately.
Include at least one newly generated article in a template regression test. A corrected older page is not proof that future pages will stop inheriting the same promotional default.
5. Report publication evidence without inventing search evidence
A defensible completion note says that the article declares the intended representative image, the asset is publicly reachable and the tested template no longer inserts the unrelated default. List any untested variants.
If a search preview is later observed, record the query, date and environment separately. Do not describe an unseen search result as repaired simply because metadata was deployed. This distinction helps publishers improve the evidence they control while being honest about the selection and timing they do not.