AEO
When ChatGPT and Gemini answer the question for your buyer, what matters is whether they mention you, and what they say.
Rankings used to be the whole story. Now the question is different: is your brand mentioned when someone asks an AI assistant the question that used to send them to a search results page, and what does the answer say about you? For a cybersecurity SaaS client, I built and ran AI sentiment analysis - batching realistic buyer queries against ChatGPT and Gemini and analysing how the brand was actually described, not just whether it turned up. As I put it to the client: it's not about where we rank, it's about how they talk about us. That's answer engine optimisation - AEO - in practice.
The pilot found the two engines disagreed. One rated the brand neutral, citing a specific perceived weakness. The other rated it positive and ranked it in the global top five. That disagreement is exactly the intelligence a content strategy needs, because you fix sentiment by building content in the areas where presence is weak, and you can't do that until you know where the weakness actually is.
Sentiment, not just presence
Most AI visibility tracking stops at "are we mentioned". That's not enough. The more useful question is what the answer says once you're in it - is the brand described accurately, favourably, or is there a specific gap in how the model perceives you. A brand that's mentioned but described unfavourably hasn't won anything by appearing. Two engines can hold two different views of the same brand at the same time, and both are worth knowing, because they point to different gaps and potentially different fixes.
Scoped from real ranking data
The method starts from keywords you already track that trigger AI results, not a blind guess at what buyers might ask. That keeps the query set realistic and the findings actionable, because they're grounded in searches you already know matter to your buyers, rather than a generic list of questions a template might suggest.
Affordable to run
This isn't a huge undertaking - around 50 well-chosen queries and a modest API budget is enough to get a real read on sentiment across the engines that matter. It's a pilot you can run before committing to anything bigger, which matters because most businesses have no baseline at all right now, and you can't fix sentiment you haven't measured.
The impressions-up, clicks-down pattern
A lot of sites are seeing impressions rise while clicks fall - a sign that AI Overviews and chat assistants are absorbing informational traffic that used to click through to the site. It's an easy pattern to misread as a ranking problem when it isn't one. The diagnostic matters because the response is specific: make the commercial pages convert properly, and work to become the cited source on the informational ones, rather than assuming the traffic will come back on its own, because in a lot of cases it won't.
Market mapping for any brand
The same method works beyond one client - a market map of who the LLMs currently think is good in your category, built from the same query-and-analyse process. It tells you where you actually stand against competitors in a channel most businesses aren't measuring at all yet. Most competitor analysis still looks at search results pages. This looks at what the AI assistant tells someone when they ask it to recommend a provider, which is quickly becoming the more consequential answer.
Most businesses have no idea what ChatGPT says about them when a buyer asks. That's not a gap you can afford much longer - the engines are already forming a view, with or without your input. Find out what it is, then decide what you're going to do about it.