Skip to main content
Cartoon robot behind a counter labelled Answer Engine Optimisation, holding up the one clearly labelled jar marked The Answer and saying here's the clearest, most-structured answer, while a stack of jars with unreadable scribbled labels sits unchosen to the side.

Answer Engine Optimisation

Answer Engine Optimisation, or AEO, is the older, plainer craft of writing your pages so they answer a real question so cleanly that a search engine can lift your words straight into the box at the top, the featured snippet, the voice assistant's reply, the AI summary, without a human ever having to click through to find them.

58%of people saw at least one AI summary in a single monthPew Research, March 2025
88%of AI summaries cite three or more sources, so there is room for more than one winnerPew Research
53%of ten-word-plus searches trigger an AI summary, versus just 8% for one or two wordsPew Research

If GEO is about getting your name mentioned inside the AI's written answer, AEO is the groundwork underneath it. It is the unglamorous business of being the answer in the first place. The two get used as if they mean the same thing, and honestly the line between them is blurry enough that I will not lose sleep if you treat them as cousins, but it is worth knowing which one is which, because AEO came first and most of what makes GEO work is just AEO done properly.

None of this is actually new

Here is the bit that surprises people. Search engines have been pulling direct answers out of websites and showing them in a box for years, the "position zero" thing, the little panel that tells you how long to boil an egg before you have clicked anything at all. Voice assistants did the same, reading one website's sentence aloud as if it were handed down on a tablet. AEO is simply the practice of being the website that gets chosen for that job. What has changed is that the box has grown teeth. It used to sit politely above the links. Now, more and more, it is the whole answer, and the links underneath are an afterthought that almost nobody reads.

How common AI answers have become

We can put numbers on this, because guessing is for mugs. A Pew Research study tracked the actual searches of around 900 American adults through March 2025, nearly 69,000 of them, and found that 18% of all Google searches now produce one of these AI summaries, and that 58% of people, more than half, ran at least one search that month that threw one up Pew Research, HIGH. So this is not some fringe feature you can fold your arms and wait out. It is most of the people you would ever want as customers, getting their answer in a box, starting roughly now.

Cartoon robot holding a jar whose label reads structured content, clean answer ready to use, with a double-headed arrow showing the robot's speech bubble quoting that same label back word for word, illustrating that an answer engine repeats your clear content rather than rewriting it.

The good news: the box has room for more than one

Now for the part that should stop you reaching for the brandy. Being the answer is not a single golden slot you have to elbow a competitor out of. When Pew looked at what was actually inside these summaries, it found that 88% of them cited three or more sources, and only 1% leaned on a single source Pew Research, HIGH. The machine is not crowning one winner and binning everyone else. It is stitching together several pages it trusts, which means there is room on the shelf for more than one jar, and one of them can be yours, even if you are a two-person firm in Frome and the others most certainly are not.

What to actually write

So how do you become one of the pages it reaches for. The honest answer is dull, and it is the same dull answer Google itself gives, which is that there is no secret switch and no magic markup that unlocks the box, just a page the search engine can read and trust that actually answers the thing being asked Google Search Central, HIGH. But the Pew data does hand you one genuinely useful steer. Short, lazy searches barely trigger these answers at all, only 8% of one or two word searches produced a summary, whereas 53% of searches with ten words or more did Pew Research, HIGH. People talk to the machine now. They type whole questions, full sentences, the way they would ask a person across a counter. Which tells you exactly what to write: the real questions your customers ask, in their own words, answered properly near the top of the page, not buried three scrolls down under a hero image and a video nobody ever pressed play on.

It helps to know the shape the machine likes, too. The typical AI summary in that same study was just 67 words long Pew Research, HIGH. Not an essay, a clear and tight and self-contained answer. So when you write a page, lead each section with the answer stated plainly in a sentence or two, then expand underneath for the people who want the detail. Answer first, working out second. It reads better for a human in a hurry, and it is exactly the chunk the machine can lift.

The bottom line

That is Answer Engine Optimisation. It is making sure that when somebody asks a question your business could answer, your page is written clearly enough, plainly enough and helpfully enough that the machine picks your words to say out loud. Get that right and GEO mostly takes care of itself, because being in the AI's answer starts with being the answer.

One small honesty note, because this is meant to be the page you can trust. Google did not love the Pew study and called the method unrepresentative, which it was always going to, and these figures will drift as the whole circus settles down. But the direction is not in dispute by anyone, and the direction is the only thing you need to act on.

If all of this sounds like a job you would rather hand to someone who does it for a living, that someone is me. Cheers, Dan.


Source confidence

ClaimConfidenceSource
18% of US Google searches triggered an AI summary; 58% of users saw at least one; ~900 adults, ~69,000 searches, March 2025HIGHPew Research, cross-verified against Search Engine Land, Search Engine Roundtable, Fortune
88% of AI summaries cited three or more sources; only 1% cited a single sourceHIGHPew Research
Just 8% of one or two word searches triggered a summary, rising to 53% for searches of ten words or more; median summary 67 wordsHIGHPew Research
No special technical setup needed for AI answers beyond standard indexing and snippet eligibilityHIGHGoogle Search Central

Page confidence: HIGH. Note for audit: the original Manus draft cited "over 2 billion monthly users" from a Semrush blog. Google's own first-party figure was about 1.5 billion (2025 earnings). The page does not lean on a raw user count at all, to avoid repeating the GEO page's scale beat. Manus's secondary Semrush stats ("88% informational", "68% of trigger terms have under 100 monthly searches") were dropped as second-hand MEDIUM and not needed for the argument. Google has publicly disputed the Pew study's methodology; the direction is unchallenged, the exact figures will move over time.

Related terms