Key Takeaways
- Write in liftable units, not long paragraphs. An assistant repeating you needs a chunk that makes sense on its own. Structure is how you hand it one.
- Give every value a label that travels with it. A price in a table keeps its column heading. The same price in a sentence loses its context the moment it is quoted.
- Turn your comparisons into actual tables. If you are already comparing options in prose, the table version is the same words in a shape a machine can lift row by row.
- Keep the prose, add the structure. Tables and lists are not a replacement for explaining yourself. They are the summary a machine reaches for after the explanation convinces it.
- Check whether your best page has any structure at all. Most coaching sites have none: no headings that ask questions, no lists, no table anywhere. That is the cheapest gap on this list to close.
Run the same question past a few assistants and watch what they cite. The sources are rarely the most beautiful pages on the internet. They are the ones shaped like reference material.
Tables get lifted. Lists get lifted. Question headings get lifted almost verbatim. Long elegant paragraphs get summarized into something that no longer mentions who wrote them.
This is a note about why that happens, what it means for your pages, and the reformatting job that costs an afternoon and no money at all.
Why does AI quote tables and lists so much?
Because both are made of small self-contained units. A table row or a list item can be pulled out on its own and still mean something, which is exactly what an assistant needs when it is assembling an answer from several sources.
A paragraph does not work that way. To use one sentence from the middle of a paragraph, the assistant has to read the whole passage, work out what the sentence depends on, and write a version in its own words. What comes out the other side is a summary, and summaries lose attribution more often than quotes do.
So the practical difference is not that machines dislike prose. It is that prose gets absorbed while structure gets repeated, and only one of those keeps your name attached.
What a machine is actually looking for
Three properties make a chunk of your page easy to reuse, and all three are formatting decisions rather than writing talent.
The first is self-containment. The unit has to survive being separated from everything around it. The second is a label that travels with the value: a price in a table keeps its column heading, while the same price in a sentence loses its context the moment it is quoted.
The third is comparability. Two rows that describe the same kind of thing let an assistant answer a comparison question directly, which is what a lot of buyer questions actually are. The GEO research paper found the same family of changes moved source visibility on its benchmark: quotations, statistics with sources, and cited references, all of which are units with their provenance attached.
None of this contradicts writing well. Google's helpful content guidance still describes pages written for people first, and a page of bare tables with no explanation reads as thin to a person and to a machine.
The same facts in three shapes
Take one set of facts, your prices and what each tier includes, and put it on the page three ways. The words barely change. What changes is how much of it can be reused without a rewrite.
In text, so the figure holds no claim of its own. The paragraph is one long unit, and to use any part of it an assistant has to read the whole thing and write its own summary. The list is several short units, each line standing alone, so one line can be lifted without the rest coming with it.
The table is units with labels attached, because every cell carries its column heading, which means a single row survives being pulled out of the page entirely. That last property is why comparison tables turn up as cited sources more than any other block we watch.
Formats compared on what a machine gets
Here is the same judgement applied across the formats a coaching site actually contains. The frequency column describes what we keep seeing in our own test runs, not a measured rate.
| Format | What a machine gets from it | Stands alone? | Turns up as a source | Best used for |
|---|---|---|---|---|
| Comparison table | Labelled values it can quote row by row | Yes | Often | Prices, formats, options, who each one suits |
| Question heading with a direct answer | A matched question and a quotable first sentence | Yes | Often | The questions buyers actually ask you |
| Numbered list of steps | An ordered procedure it can restate | Yes | Often | How something is done, in order |
| Bulleted list | Separate short claims | Mostly | Sometimes | Inclusions, requirements, what is not covered |
| Explanatory prose | Understanding it will summarize in its own words | No | Sometimes | Convincing a reader, which still matters most |
| Testimonial slider | Often nothing, if the text loads after the page does | No | Rarely | Reassuring visitors, not being quoted |
| Text inside an image or video | Nothing at all | No | Rarely | Anything you do not need repeated |
The bottom two rows are where most coaching proof currently lives, which is the uncomfortable part. A results carousel and a screenshot of a happy message are both invisible to the thing writing the answer.
The pattern across page types, rather than block types, is in what kinds of pages assistants quote most, and the structured markup question, which is a different one, is covered by the schema.org FAQPage definition if you want the technical version.
How to reformat what you already have
You almost certainly do not need new content. You need the content you have, in a different shape, which is an afternoon of work.
- Find the comparison hiding in your prose. Anywhere you wrote "unlike", "whereas", or "the difference is", there is a table waiting. Two columns and three rows beats four paragraphs.
- Turn your headings into questions. Replace "Our Approach" with the question a buyer would type, then answer it in the first sentence underneath.
- Put your prices in text. A table of what each tier costs and who it suits is quotable. A price behind a form is not, and an assistant will happily quote a competitor who published theirs.
- Lift the text out of your images. Anything that currently only exists inside a graphic or a video needs a text version on the same page. The plain-text version of client results is the highest value example.
- Give each list item a whole sentence. Fragments that only make sense in sequence are not liftable. If an item cannot survive being quoted alone, rewrite it until it can.
The full structure for the single page most likely to earn a citation is in how to write a comparison page AI actually quotes. Do that page first, then work backwards through the rest of your site.
Then measure rather than assume. Ask twenty-five buyer questions, look at which sources come back, and see whether any of them are yours. Our free check runs exactly that count at $0.
Questions we hear the most
Why do AI answers quote tables and lists so often?
Because both are made of small self-contained units. A table row or a list item can be lifted out and still make sense, while a paragraph has to be read whole and rewritten before any part of it can be used.
Does adding a table to my page really help?
It helps when the table holds real comparable values. A genuine table of prices, formats, or options gives an assistant something to quote directly, while a table of vague marketing phrases gives it nothing new.
Should I turn all my paragraphs into bullet points?
No. Explanation still needs prose, and pages that are nothing but bullets read as thin to people and to machines. Write the explanation, then add the structured summary underneath it.
How long should a list item be?
Long enough to stand alone, short enough to quote. One clear sentence usually does it, and an item that only makes sense after reading the item above it is not really a list item.
Do headings matter as much as tables?
Almost. A heading phrased as the question a buyer asks, answered in the first sentence beneath it, is the single most quotable pattern we see, and it costs nothing to adopt.
Does this mean I should add schema markup to my tables?
Not for tables specifically. Structured markup helps for questions, reviews, and business details, but the plain HTML table and list are already readable, so write those first and treat markup as a later refinement.
How do I find out which of my pages AI already quotes?
Ask an assistant a question your page answers and look at the sources it cites. Repeat across twenty-five buyer questions and you will see which pages of yours, if any, are being used.
What does the AnswerHalo free check include?
It costs $0. We ask ChatGPT 25 real buyer questions from your niche, then email a plain-English readout within the hour: how many answers named you, who was named instead, and the first fix. It runs as soon as you ask for it.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →