AI-Written Alt Text and Image Captions for Affiliate Guides: What to Generate and What to Write Yourself
Alt text is the part of a guide nobody reads, written for the readers who need it most, and on most affiliate sites it says "image-1" or "best bike lights for commuting" twelve times.
Both are failures: the first tells a screen-reader user nothing, the second tells Google the page is stuffing keywords. The honest version is short, specific and easy to get a tool to write, as long as you give it the one thing it can't guess.
This post covers what alt text is for and what Google does with it, why generated alt text defaults to the filename or the keyword, the formula that works for product images and hero images, where captions earn their place and what they should say, which of the three jobs to leave to a tool and which to do yourself, and a thirty-second accessibility check.
The images post covers which product images you're allowed to show; this is what to say about them.
The short answer
Alt text describes what the image shows, for someone who can't see it: product name plus the visible thing, under about 125 characters, no "image of", no keyword list.
Hero images get a sentence about the scene; product images get the product and what the picture makes visible; decorative images get empty alt. Captions go on the hero and the top pick only, and say a reason, not a name. Let a tool draft; read every one; the usual failure is repetition.
What alt text is for
Two readers. A screen-reader user hears the alt text in place of the image, so it's the only version of the picture they get; on a buyer guide that means the difference between "there's a product here" and knowing which product and what it looks like mounted.
Google reads alt text to understand the image for image search and as one of many signals about the page's subject. The version that serves the first reader serves the second; there isn't a separate "SEO alt text", and writing one is how pages end up with twelve identical keyword strings.
Why generated alt text goes wrong
- The filename default. The tool has no image understanding and copies "product-3.jpg". Common in bulk-publishing tools.
- The keyword default. The tool knows the post's keyword and puts it on every image. Reads as stuffing; tells a screen reader nothing.
- The generic caption. "A black bicycle light on a white background." Accurate, useless; which light?
- The missing product. Image-recognition tools describe what they see and can't know the model name. They need to be told which product the image belongs to.
The fix for all four is the same input: the product name attached to each image. A tool that has it writes "Cygolite Metro Pro 1100, side view showing the handlebar mount"; a tool that doesn't writes one of the four above.
The formula
| Image | Alt text pattern | Example | Not |
|---|---|---|---|
| Hero (the post's lead image) | What the scene shows, one sentence, product if one is featured | "Commuter bike at dusk with a front light lit, city street behind" | "Best bike lights for commuting" |
| Product image (API-served) | Product name + what's visible | "Lezyne Micro Drive 800XL, front view, strap mount visible" | "Product image" / the filename |
| Comparison or detail shot | What's being compared or shown | "Beam pattern of the Metro Pro versus the Micro Drive on a dark road" | The keyword |
| Decorative (dividers, icons, backgrounds) | Empty: alt="" | alt="" | "Decorative image" |
| Author photo | The person's name | "Sam Ortiz" | "Author" |
Under about 125 characters, because some screen readers cut off there. No "image of" or "picture of"; the reader already knows it's an image. Punctuation helps: a comma is a pause when read aloud.
Captions
Alt text is invisible; captions are read by everyone and sit exactly where the eye lands. That makes a caption the best place on the page for one short reason: "The Metro's mount is the one that survives a winter" under the top pick's image does what a body sentence does, in the spot readers actually look. A caption that repeats the product name wastes the spot.
Use captions on the hero (what this guide is for, in one line) and on the top pick (the reason). Skip them on the rest of the products; a page with ten captions reads as a catalogue. The opinions post covers where the reason comes from; the caption is one of the places it goes.
What to generate and what to write
- Generate: product-image alt text, given the product name per image. It's formulaic and a tool with the name does it correctly.
- Generate, then read: hero alt text. Tools describe scenes well; check it isn't the keyword in disguise.
- Write yourself: the two captions. They carry the opinion, and the opinion is yours.
- Set once: decorative images to empty alt in the theme, so they stop being "divider.png" to a screen reader.
The thirty-second check
- Open the guide, view the page outline or use a browser extension that lists alt text (most accessibility checkers do).
- Read the alts in order. If two are identical, one is wrong. If any is the keyword, it's wrong. If any is a filename, it's wrong.
- Confirm decorative images read as empty, not as a filename.
- Read the two captions aloud. If either could be deleted without losing a reason, rewrite it.
The editing post covers the rest of the fifteen-minute pass; this is thirty seconds of it that most sites skip.
Doing it with AffSERP
Guides written by AffSERP's X Best Review tool carry the product name on every API-served product image, so the alt text follows the formula above rather than the filename, and the hero image gets a scene description rather than the keyword.
Decorative elements ship with empty alt. The two captions are yours: the tool leaves the hero and top-pick caption fields for the reason you'd give a friend, which is the one thing no formula produces.
FAQ
What should alt text say for affiliate product images?
What the image shows, for someone who can't see it: the product name and the thing the picture makes visible. 'Cygolite Metro Pro 1100 mounted on a handlebar, beam on' tells a screen-reader user what's there; 'best bike lights for commuting 2026' tells them nothing and reads as keyword stuffing to Google. Product, what's shown, under about 125 characters, no 'image of'.
Can AI generate alt text?
Yes, and most tools default to one of two bad outputs: the filename ('IMG_4821' or 'product-1') or the post's keyword repeated on every image. A tool that looks at the image and writes what's in it does better, and a tool that knows which product the image belongs to does better still. Either way, read the result; the usual failure is ten images with the same alt.
Does alt text help SEO on an affiliate site?
Modestly: it's how Google understands what an image is, which matters for image search and as one signal of what the page is about. It won't move a guide's ranking on its own. Its main job is accessibility, and the version written for accessibility is also the version Google rewards; the version written for Google (keyword, keyword, keyword) fails both.
Should product images have captions?
Captions are optional and they're read far more than alt text. A caption that adds a reason ('The Metro's mount is the one that survives winter') does the work of a sentence in the body, sits next to the picture where readers look, and is visible to everyone. A caption that repeats the product name does nothing. Use them on the hero and on the top pick; skip them elsewhere.
Keep reading
