In the past, a brand that wanted to be remembered would run TV ads, buy Yellow Pages listings, place ads in newspapers and magazines, or put up a huge billboard on the street. That was an era of “if I show up often enough, you’ll remember me”.
Then the internet arrived: Google Search, Facebook, Instagram, YouTube, one after another. Brands started studying keywords, rankings, traffic and click-through rates. Everyone was still competing to be seen; the battlefield just moved from TV, newspapers and the street onto screens. As we discussed before, the entry point has shifted from Google Search to AI conversations, but being remembered has never depended on the entry point. It depends on the content itself.
But since AI appeared, I think things have changed again. A customer may not have searched for any brand at all. They simply open ChatGPT or Gemini and ask: “Which companies offer this kind of service?” And the AI puts together an answer for them first.
AI has started building shortlists for customers
Recently I keep seeing this scenario in corporate procurement. The person responsible for finding vendors asks ChatGPT before a meeting, “Which companies would you recommend for this kind of service?”, gets a list, and only then starts researching them one by one. Or a manager comparing two consultancies drops both websites into Gemini and asks it to summarise the differences.
We used to make comparison tables ourselves. Now AI makes them for the customer.
There is an even quieter situation: the brand doesn’t even know it was asked about. No ad impression, no form, perhaps not even a single website visit. A comparison involving you has already taken place, and you don’t even know whether you made the list.
This is a change I’ve felt strongly recently.
We used to worry that our ranking wasn’t high enough. Now we may first need to worry that we weren’t counted in the answer at all.
In G2’s 2026 survey of more than 1,000 B2B software buyers and decision-makers, 71% said they use AI chat tools during software research, and 69% had ended up choosing a different vendor from the one they originally expected because of information provided by AI. More interesting still, 33% ended up buying from a company they had never heard of before AI recommended it.
Of course, this is market research on B2B software buyers. I won’t apply it directly to every industry, and I don’t think a few percentages are enough to prove that everyone’s buying behaviour has changed. But the direction is worth watching, because AI is no longer just helping people “find information”. It is starting to help people organise options, compare differences and even form shortlists.
So recently I’ve been asking brands a question I rarely asked before:
When a customer doesn’t type in your brand name and just asks AI “who should I go to in this field?”, will it think of you?
Google added a small button
A very small but, I think, noteworthy signal appeared recently. On 20 August, Google updated its official documentation for Preferred Sources, adding an interactive button that websites can place on their own pages.
How it works is simple. A website adds the code Google provides to its page. When a reader clicks it, they can add the website to their Google Search preferred sources, and are then returned to the page they were reading.
Preferred Sources itself did not appear on 20 August. It started with Top Stories, and in May this year Google extended it to AI Overviews and AI Mode. Once a user sets a website as a preferred source, that site’s content has a better chance of appearing in the Top Stories this user sees, and in AI Overviews and AI Mode it is also identified with a preferred source label.
There are two things here that I think need to be made especially clear.
The first is the numbers
Google itself has said that after users add a website to Preferred Sources, they are on average about twice as likely to click through to that site. But this “twice” is not a performance figure promised in Search Central technical documentation. It is an observation Google published on its own product blog.
So the direction is worth noting, but I won’t translate it directly into:
“Install this button and your traffic will double.”
Those are two very different things.
The second is more important
This is not a magic SEO button.
It affects “the one person who clicks it”. In other words, if a reader actively sets CloudAD as their Preferred Source, it is that reader who is more likely to see CloudAD content in their own Google Search, AI Overviews or AI Mode in future. It does not mean the whole CloudAD website ranks higher, and installing more buttons does not mean traffic from strangers will flow in on its own.
So if you look at it purely as a tool for acquiring new traffic, I’m not that excited about it yet.
But I installed it anyway.
We don’t know, which is why it’s worth trying
In the past few days CloudAD has placed this button at the end of the body text on the article pages of our website. Why? Because I have always felt that when something new comes along, rather than first writing ten articles telling everyone how important it is, it’s better to try it yourself first.
If it’s worth trying, try it first. If it can be tested, keep a record of the test. If we don’t know yet, don’t rush to a conclusion.
Since installing it, there actually isn’t much to see yet, because Preferred Sources is a setting in the user’s personal account, and the feedback a website can receive is very limited.
So I can’t tell you, “We installed it three days ago and the results are great.”
No. Right now we simply don’t know. And I actually think this “not knowing” is well worth recording.
An experiment is not meant to prove you’re right. It’s meant to help you gradually know a bit more about something you didn’t know.
That’s also why I want to follow this small button. What really interests me is not how much traffic it can bring right now, but that Google was willing to create a physical entry point that lets users actively tell the system:
“This is a source I want to keep seeing.”
Platforms used to guess what we wanted to see. Now Google is starting to let users tell it directly: “I want to keep seeing this source.” That action alone is worth watching.
So we tested it on CloudAD
So we also ran an experiment on CloudAD itself. During our recent full-site health check, we deliberately tested a set of questions closer to real customer decision scenarios. Not searching for “CloudAD”, but asking questions such as:
“Which GA4 consultancies are there in Taiwan?”
“Who can a company go to for GA360?”
“Which MarTech data consultancies would you recommend?”
and so on.
The results were interesting, and very honest:
AI doesn’t yet tend to mention CloudAD on its own.
I wasn’t surprised at all. When we ran our website in the past, we were mostly thinking about how people search, how Google searches and how keywords rank. We had never deliberately re-examined ourselves from the angle of “if an AI were trying to understand CloudAD today, would it actually know who we are?”
I believe CloudAD isn’t an exception. Many capable small and medium-sized businesses may be in a similar position: they’ve provided their services for years, have plenty of clients and have kept building up website content, but before AI arrived we never had a reason to check:
whether machines have actually pieced all this scattered content together into a clear picture of “who you are”.
Content being well written is one thing. Being understood by the system is another. And once understood, whether it’s willing to mention you when someone actually asks “who should I go to?” is a third.
I really want to look at these three things separately now. Not because I already know the answer. Quite the opposite:
it’s precisely because we don’t know that it’s worth trying.
Next, we will keep recording CloudAD’s visibility across different AI tools, different questions and different points in time, to see whether the answers change as website content, brand information, external citations and source signals change.
Maybe it will work. Maybe some of it will be completely useless. Maybe six months from now we’ll overturn half of what we think today. That’s all fine.
At least it will be an answer we arrived at after running through it ourselves, rather than quoting yet another “Top 10 GEO trends for 2026”.
Go and ask first
So if you ask me now what brands should actually do in the AI era, I won’t rush to give a standard answer. I’ll start with a very simple question:
When someone who knows nothing about you asks AI “who should I go to for this?”, is your name counted in the answer?
Go and ask. Try a few different phrasings, ask ChatGPT once and Gemini once, and ask again after a while. See how it describes you, see who it puts next to you, and see whether it mentions you at all.
There is now a way of looking at whether a brand is seen that didn’t exist before. Where you rank in search engines still matters, but beyond that we may also need to start observing:
whether AI actually knows you, and which version of you it knows.
We’ve only just started experimenting with this. If you’d like to talk about it, clouda now has a new option:
“Not sure whether AI mentions us when it answers questions.”
Clicking it won’t start with a promise of any ranking, and it won’t tell you there’s a magic formula. We can first look together at how AI currently understands your brand, what it cites and what it doesn’t see. After all, we’re still figuring this out ourselves, and right now no one can guarantee how the black box chooses.
As for the Preferred Sources button we installed, the AI visibility tests, and whatever we observe next, I’ll keep recording the results.
When there are results, I’ll come back and share them.
Talk to clouda and see whether your brand information is clear enough
