"People Also Ask" is one of our secret weapons in AEO. Let me explain. If you have not looked at it closely, it’s the box of related questions Google shows midway down the results page. Click one and you get a short answer. We like it for two reasons. Firstly, these are real questions that real customers are asking, so nobody has to guess at them in a brainstorm. Secondly, they are written the way customers talk. That is also how people type into ChatGPT, Claude or Perplexity, which makes it a useful preview of what you will be asked in AI search. We ran this for one of our law firm clients: - 168 seed searches based on their buyer questions - 1,083 unique People Also Ask questions came back - 441 were relevant enough to keep as content signals Some of what came back: - "Is 50k a good settlement for a car accident?" - "Will my insurance go up if I'm not at fault?" - "Should you give your lawyer your Social Security number?" - "What scares insurance adjusters?" As far as data can tell, not a single person searches for "comprehensive legal representation tailored to your needs." Instead, they ask whether the offer sitting on their kitchen table is a fair one. Here’s the process we follow: 1. List the questions your buyers ask on their way to a purchase. 2. Google each one and note which People Also Ask questions appear. 3. Then check whether you or your competitors are the ones answering them. 4. Use every gap to map to a new page or an improvement to one you already have. Hatter automates this by continuously monitoring People Also Ask as one of our signals and surfacing them as opportunities for you to act on. Customers have been telling us what they want to know for years. The work is mostly in taking the time to listen.

Is being recommended in AI assistants a volume game? Something we noticed this year has changed the way we think about that question. When we look at which pages AI assistants actually cite, the deciding factor is rarely how much a business has published. It is whether a page was written for someone who is close to making a decision. AI assistants, it turns out, are not especially impressed by the amount of effort. Here is what the last 30 days looked like for two of our clients. A marketplace with over 400 blog posts on Google: - Four of its most-cited blog posts answer questions people ask right before they book or list, such as how it works. Together, they earned 60 AI citations. - Its two most popular blog posts on Google, which brought in over 14,000 clicks in 90 days, did not make its list of 20 most-cited pages. A law firm with over 400 pages on Google: - Every page AI assistants cited was written for someone looking for a lawyer, including articles on specific situations, specific cities, and how to choose the right firm. - Its highest-traffic blog posts, which cover general car problems, were not cited once. Those traffic posts are still doing their job on Google, and we would not remove them. But they show that being useful to a reader is not the same as being useful to a buyer, and AI assistants seem to know the difference. This is somewhat inconvenient for an industry whose answer to AI search has mostly been to publish more, since "more" is considerably easier to sell than "better." That is why we recommend no more than five pieces per week. Each page should be written for a specific moment in the buyer's journey, and doing that well takes deliberate work. It means understanding what the buyer is thinking at that stage, what they need to know, and what they require before they feel confident moving forward. Only then can you decide which page should answer them. Five pieces a week is roughly what that level of care allows, and it has the pleasant side effect of protecting your site health. In our experience, the businesses that do best in AI search are not the ones that say the most, but the ones that say the right things clearly.

Claude Fable 5 is the first model to unlock a self running agentic loop in AEO. It's cooking...