Should important facts appear in image text or body copy
Put them in body copy first, then let images support the point

Why is body copy safer than image text for important facts?

Because extraction starts with what the system can fetch, parse, and segment reliably. Plain HTML copy gives crawlers and downstream answer systems a direct path to the words. Text baked into an image adds another layer of failure. The system has to notice the image, process it, read it correctly, and connect that text to the surrounding claim.

That may sound obvious, but teams still hide key facts inside product screenshots, comparison graphics, hero banners, and infographic blocks. It looks polished. It often performs worse for retrieval.

There is a more basic reason too. AI crawlers do not execute JavaScript, they fetch JavaScript files and never run them. That verified behavior matters here because many modern sites already make extraction harder than it needs to be. If your page relies on client side rendering and your only clear fact statement is also trapped inside an image, you are stacking two avoidable problems on top of each other.

• Body copy is directly parseable from HTML.
• Sentences in HTML are easier to quote accurately than text embedded in graphics.
• Lists and tables give models clearer boundaries around facts.
• Images can be skipped, misread, or detached from context.
• If the page is JS heavy, image only facts become even less dependable.

When does image text still help?

Image text still helps people. It can also help machines in some cases, especially when the image is central to the page and the surrounding copy explains what the reader is seeing. But support is not the same as source of record.

I like image text for three jobs. First, to make a proof point feel concrete, such as a screenshot of a settings panel or report. Second, to speed scanning, such as a labeled process diagram. Third, to add persuasion after the factual claim has already been stated in normal copy.

The rule is simple. If you would be annoyed to see the fact omitted from an AI answer, do not leave it inside the image alone.

A practical standard I use

Any fact that matters to decision making should appear in a complete sentence in body copy. If it compares options, it should also appear in a table. If it defines a process, it should appear in a list or short steps. Then the image can visualize the same point without carrying the whole burden.

What actually goes wrong when facts live only in images?

A few different failure modes show up. The first is simple omission. The assistant answers the question from surrounding copy and ignores the image entirely. The second is partial extraction. It catches a label but misses the qualifier, timeframe, or condition attached to it. The third is context drift. The image says one thing, the nearby paragraph implies another, and the answer engine chooses the cleaner sentence rather than the graphic.

This is why operator style pages tend to outperform pretty pages for citation. They are easier to parse. They say the important thing plainly, close to the top, with less decorative friction.

If you want a page to survive summarization, write as if a rushed researcher will pull one paragraph and one table row. That is usually close to what happens.

Should you repeat the same fact in text, alt text, and captions?

Yes, with discipline. Repetition helps only when each location has a distinct job. Body copy should carry the factual statement. The caption should explain what the image demonstrates. Alt text should describe the image for accessibility, not serve as your hidden fact dump.

People get this wrong by treating alt text like a side channel for SEO or AI. It is not a substitute for visible copy. If the key claim matters, say it where readers and machines can both see it.

I would rather see a plain sentence above an average screenshot than a brilliant graphic with no extractable explanation around it.

A clean pattern

• Lead with the claim in one sentence.
• Place the image immediately below or beside it.
• Add a caption that states what the image confirms.
• If the fact compares options, add a table row with the same point.
• Keep wording consistent across placements so assistants do not face contradictory phrasing.

What kinds of facts should never live only in images?

Definitions, product capabilities, exclusions, methodology notes, dates of change, and anything with a trade off should never live only in image text. These are exactly the details that get lost when systems summarize quickly.

The deprecation of FAQ rich results is a useful reminder here. Google fully deprecated them on 2026-05-07. That means publishers cannot lean on old presentation habits and expect surface level formatting tricks to carry visibility. The underlying page still has to communicate clearly in visible copy. Good extraction now depends more on page clarity than on hoping a special treatment will rescue weak structure.

The same caution applies to llms.txt. Google states it is not used by Search. Separate research across a large domain set found low adoption and no citation lift after controls. So if a team asks whether an llms.txt file can compensate for hiding core facts inside visuals, the answer is no. Put the facts on the page in normal text first.

Where does this advice fail or become less important?

It matters less when the image itself is the object of interest rather than the text inside it. A product photo, UI layout, packaging design, or chart shape can still contribute value even if the system does not read every label. In those cases the surrounding explanation carries the extractable meaning, and the image carries the visual evidence.

It also matters less on pages that are not trying to win retrieval or citation. Brand campaigns, visual storytelling pieces, and portfolio pages can accept more ambiguity. Not every page needs to be tuned for answer engines.

And here is the honest limitation. Putting facts in body copy does not guarantee citation. If your site lacks authority, if stronger third party sources phrase the same claim more clearly, or if your own pages contradict each other, plain text alone will not save you. This is a necessary condition for extractability, not a complete visibility strategy.

Who should not follow this advice too rigidly? Design teams building image led editorial experiences. Do not flatten everything into ugly walls of text just to satisfy a retrieval theory. The move is not less design. The move is clear factual copy paired with design that supports it.

If your main problem is not content structure but demand generation execution, that belongs with the parent brand. We run managed outbound under Outbound Pros, but outbound mechanics are a separate system from AI retrieval. Keep the two disciplines connected, not confused.

How should you format body copy so assistants can lift it cleanly?

Use short declarative sentences. Put the claim before the explanation. Keep qualifiers close to the fact they modify. Avoid burying the answer inside metaphors, slogans, or long scene setting intros.

For important operational pages, I like this sequence. Start with a direct answer. Follow with a paragraph that explains why. Then add a list or table for the edge cases. This pattern gives searchers what they need and gives retrieval systems neat extractable chunks.

• State one key fact per sentence when possible.
• Put the most important fact high on the page.
• Use headings that match the question a buyer would ask.
• Keep supporting visuals close to the relevant paragraph.
• If a chart matters, summarize its takeaway in text immediately above or below it.
• Do not force readers to infer the claim from a screenshot.

This is not glamorous advice. It is the sort of thing teams skip because everyone wants a bigger lever. In practice, pages that get cited usually remove friction before they add flair.