Good context is where AEO starts.

The usual list is to add structured data, write in questions and answers, use clear headings and keep pages fast. All of that helps.

But for a lot of companies the facts underneath are not consistent, current or complete, and that is often the bigger reason assistants describe them badly.

1. How an answer engine answers

When someone asks an assistant about your company, many assistants now search the web before they reply. They pull a handful of pages onto the desk, meaning into the context the model is working from, and write the answer from what those pages say. The model’s training still plays a part, but for anything specific and current, the pages it just found carry most of the weight.

Seen that way, AEO is the work of influencing which pages land on that desk and what they say once they get there. It is a context problem, played out on a desk you do not own.

2. The answer is becoming the destination

This matters because fewer people click through to check. Pew Research Center tracked the real browsing of 900 US adults across 68,879 Google searches in March 2025.

When a summary written by AI appeared, people clicked a traditional result in 8% of visits, against 15% when none appeared. They clicked a link inside the summary itself in 1% of visits.

That is an association, not proof of cause, but the direction is clear enough: for a lot of searches, the summary is the last thing the person reads.

The same study shows which searches get a summary in the first place, and they look a lot like the questions buyers ask about a company. Searches that began with who, what, when or why produced one 60% of the time, and searches of ten words or more 53% of the time, against 8% for searches of one or two words.

“Is this vendor secure?” and “how much does this cost for a team of fifty?” are exactly the kind of search that now gets answered on the page.

3. What goes wrong on your own pages

Three problems come up again and again, and none of them is fixed by formatting.

The first is contradiction: the pricing page says one thing, a help article written before the last change says another, and a partner listing says a third. An answer engine that finds all three has to pick one or blend them, and it has no reliable way to know which is current.

The second is staleness. Old pages stay indexed long after they stop being true, and in traditional search a well-linked page from two years ago can easily outrank the one you updated last week.

Assistants lean the other way. When Ahrefs analysed almost 17 million cited URLs in 2025, pages cited by assistants averaged 1,064 days old, against 1,432 days for Google’s organic results, which makes them about a quarter fresher. ChatGPT leaned furthest towards recent pages, while Google’s own AI Overviews behaved much like its organic results.

Two things follow from that. A current, dated page has a real advantage with most assistants, which is good news if your own pages are kept up to date. And the average cited page is still close to three years old, so an old page nobody has corrected can keep being quoted for a long time.

The third is absence. Buyers ask when a company was founded, whether it is secure and how it compares with the alternatives. If your own domain has no clear answer, the engine answers from whatever else it can find, which is often a review site, a forum thread or a competitor’s comparison page.

The Context Diagnostic scores this as evidence gaps: claims with nothing on the brand’s own domain behind them.

4. Why the context has to come first

Every page on your site was written by someone, and more and more of it is drafted with an assistant. If those tools work from whatever turned up, they carry the contradictions into every new page, help article and press release. Publishing more simply spreads the problem further.

A record changes that at the source. When each fact lives in one place, with a date, a source and an approver, every tool your team uses to write pages reads the same version. The pricing page, the help centre and the press release state the same price in the same way, and all change on the same day when it does.

That consistency matters to an answer engine for a simple reason. When several of your pages agree, there is much less for it to resolve, and your own account of the facts becomes the easiest one to repeat. Specific, dated statements are also easier to quote than vague ones, which is where most formatting advice is trying to get to anyway.

5. A practical order of work

1. Ask the questions your buyers ask, across the assistants they use, and keep the answers dated and word for word.
2. Trace each wrong or thin answer back to the page it came from.
3. Write the correct version of each fact into the record once, with its source and an owner.
4. Publish those facts plainly on your own domain, with a clear page for each question buyers actually ask, and retire or redirect the pages that contradict them.
5. Ask the same questions again on a schedule, so you can see which answers moved and which sources are still being used.

Formatting work, such as structured data and headings written as questions, fits into the fourth step. It works far better once the facts underneath it agree.

6. What this can and cannot do

No one controls what a stranger’s assistant says, and anyone promising a particular score or ranking is promising something they cannot deliver. What good context gives you is a better chance: your own pages become the most consistent, current and specific source available about you, which is the kind of source an answer engine is more likely to draw on.

The measurable part is the before and after. You can see what the assistants said, trace where it came from, fix the source and check again.

Start with the diagnostic. It is free, it is one pass, and you keep what it finds whether or not anything follows it.