Should the logo go on a blog featured image?

Usually a small mark in a corner on the social and Open Graph version, so a shared link is recognizably yours in a feed; often none on the version displayed at the top of the post if the site's header already carries the logo. Either way the logo is never drawn by the model. It is composited from the real file after generation, at a consistent corner and size.

← All posts

Automatic Featured Images for WordPress Posts That Look Like Your Brand, Not Like a Plugin Made Them

Q: Why do AI-generated featured images all look the same?

Because the plugins that make them generate from the post title with a generic prompt and no brand input, so the model reaches for the most probable illustration of the title: the same isometric laptop, the same gradient, the same smiling stock person. Nothing in the process carried your palette, your typography or your logo, so nothing in the output does. The fix is to give the generator those as locked assets, and to apply the ones a model cannot render (the logo, exact colors, real fonts) after generation.

Q: What size should a WordPress featured image be?

1200 by 630 pixels is the safe default because it is the Open Graph size that social platforms and messaging apps use for link previews, and most themes display it well. Generate at that ratio rather than cropping a square, keep the subject and any logo inside the central safe zone so platform crops do not cut them, and check the theme's own display size if it differs.

Q: Should featured images have alt text and a caption?

Yes, and they usually do not, because in a plugin workflow the alt text lives in the media library where nobody goes back to write it. Alt text is an accessibility requirement and an SEO signal; the caption is optional but useful. Both should be written at the same time as the image, from the post's brief, and set on the post when it is published rather than left for later.

Scroll a dozen B2B blogs that adopted AI featured images in the last year and you can spot them from across the room. The isometric laptop with floating icons. The gradient that fades from purple to teal. The smiling person at a desk who has never worked anywhere. Each is technically a unique image, and every one looks like the same image, because they were all produced the same way: a plugin took the post title, wrapped it in a generic prompt, and asked a model for an illustration.

The plugin also, in most cases, left the alt text empty, because alt text lives in the media library and nobody went back.

This post is about making the featured image yours: the three locks that separate a brand’s image from a plugin’s, the size and safe-zone rules that make it work as a link preview, and doing the alt text and caption at the same time as the image and setting all of it on the post at publish, instead of in three places afterward.

Why plugin featured images look like plugin featured images

An image model produces the most probable image for the prompt. A prompt built from a post title, “5 Ways to Reduce Onboarding Time,” has one most probable image, and it is the isometric laptop, because that is what illustrations of software articles look like in the training data. The plugin adds nothing to steer it: no palette, no typography, no logo, no reference to your last fifty images. So the output is a generic illustration of the topic, and generic is exactly what a brand is not.

Three things are missing, and they map to the three things a model cannot be trusted with anyway.

The logo. Not in the prompt, and if it were, the model would draw an approximation.
The colors and fonts. “Use our brand blue” in a prompt produces a blue. Your blue is a hex value, and the model does not do hex.
Continuity. Nothing connects this image to the last one. Each post’s hero is a fresh draw from the distribution, which is why a blog’s archive page looks like a stock library.

The three locks

A featured image is yours when three things are locked, and each is applied outside the model.

Lock 1: the logo is composited from the real file. The model generates the scene with no logo in the prompt at all. The real logo file, transparent PNG or SVG, is placed after generation: a small mark in a consistent corner, sized to the image’s short side, with a contrast pad when the background needs one. The placement guide has the numbers. For the featured image specifically, the mark usually belongs on the social and Open Graph version, so a shared link is recognizably yours in a feed, and can be skipped on the in-post version if the site header already carries the logo.

Lock 2: colors and fonts are applied as assets. The brand palette is a set of hex values and the brand typeface is a font file. Both are applied after generation: a color treatment or overlay that pulls the image into the palette, and any text on the image (a title band, a category label) set in the real font as live type. The model is prompted for mood and subject; the brand system is applied to the result. This is what makes fifty featured images look like one blog: the same palette and type on every one, regardless of what the scene is.

Lock 3: alt text and caption are written with the image, from the brief. The post’s brief already contains the audience, the keyword and the angle. The alt text describes what the image shows for a reader who cannot see it, and the caption, if used, connects the image to the post. Both are written at the moment the image is made, while the brief is open, not later in the media library. This is the lock that gets skipped most, and it is the one search engines and screen readers actually read.

Continuity across the archive. A featured image is one of a series. The campaign consistency post explains the continuity sheet: lock the environment, the lighting recipe and the palette as references, and generate every image from them rather than from the title alone. A blog whose heroes share a lighting recipe and a palette reads as one publication. A blog whose heroes share a plugin reads as a plugin.

The 1200 by 630 rule

The featured image does two jobs: it sits at the top of the post, and it is the Open Graph image that Slack, LinkedIn, X, iMessage and every link-preview renderer shows when someone shares the URL. The second job is why the size matters.

Generate at 1200 by 630, or at that ratio. Not a square cropped later; the crop is where the subject and the logo get cut. Most themes display 1200 by 630 well at the top of the post, and if yours needs a different size for the in-post display, generate for Open Graph and let the theme scale.
Keep the subject and the logo in the central safe zone. Platforms crop previews differently; a subject near an edge disappears on one of them. The QA checklist has the placement layer.
No text the preview needs to be legible. At preview size in a feed, image text under about 40 pixels tall is noise. If the image carries a title band, it is for the in-post display, and the preview has to work without reading it.
Check the Open Graph tags carry it. The featured image and the og:image are not automatically the same on every theme and SEO plugin. Confirm once.

Set it on the post at publish, not in three places after

In the plugin workflow the featured image is a media library upload, the alt text is a field on that upload, the caption is another field, and the og:image is set by the SEO plugin from the featured image if the configuration is right. Four places, visited after the post is written, which is why alt text is empty on most blogs.

The better workflow makes the image, the alt text and the caption properties of the post from the moment the image exists, and writes them to WordPress together at publish. The brief-to-published post covers the full field set; the image is one row of the eleven, and the same rule applies: set it once where it is born, carry it forward. Over the WordPress REST API with an Application Password, a publish can upload the media item, set its alt text and caption, attach it as the post’s featured image, and schedule the post in one action, with nothing installed on the site.

One more rule, and it is the one that keeps this honest. Make the image for approved copy. The headline drives the image; if the image is made before the headline is final, every headline change orphans it. Gate the copy, then make the image, once.

What this looks like built

Marqeable’s blog workflow does the three locks by construction. The featured image is generated for the approved post, one image at a time, from the brief and the brand look; the real logo file is composited after generation with placement by corner busyness and sizing by short side; brand colors and fonts come from the brand look as locked assets rather than prompt words; alt text and caption are written with the image from the brief; and the publish sets the image, alt text and caption on the WordPress post along with the rest of its fields and the schedule. The archive page looks like one blog because every hero was made inside the same harness, and the alt text is never empty because it was never a separate step.

Frequently asked questions

Why do AI-generated featured images all look the same?

Because plugins generate from the title with a generic prompt and no brand input, so the model produces the most probable illustration of the topic. Nothing carried your palette, type or logo in, so nothing carries them out.

What size should a WordPress featured image be?

1200 by 630 as the default, generated at that ratio rather than cropped, with the subject and logo inside the central safe zone so platform previews do not cut them.

Should featured images have alt text and a caption?

Yes. Write both with the image, from the brief, and set them on the post at publish rather than in the media library afterward. Alt text is an accessibility requirement and an SEO signal.

Should the logo go on a featured image?

A small composited mark on the social and Open Graph version so shared links are recognizably yours; often none on the in-post version if the header carries the logo. Never drawn by the model.

The bottom line

A featured image made by a plugin from the title is a generic illustration of a topic, with empty alt text, in a media library. A featured image made for a brand is the same generator with three locks applied outside the model: the real logo composited, the real colors and fonts applied as assets, and the alt text and caption written with the image from the brief. Generate at 1200 by 630 with the subject in the safe zone, make it for approved copy, and set the image and its text on the post at publish. The heroes then look like one blog, and shared links look like you.

Marqeable runs your campaigns, answers every visitor, text, and email in seconds, and turns them into booked jobs and meetings - even at 9pm on a Saturday. We’re in private beta with a small early cohort. Get early access