Starting today, users in the United States who joined the Search Labs waitlist can try Google's Search Generative Experience, which Google presented at I/O two weeks ago. When a user types a question, a block of AI-written text answers it above the usual blue links, with a few source links placed to the side.
It is an experiment, limited to the United States, to English and to the waitlist. Most people in Quebec will not see it for some time. However, SEO professionals reviewing the screenshots have reason for concern.
The central question is this: if Google answers the question directly, what incentive remains for users to click?
What does Google's Search Generative Experience actually show?
Google's own post, Supercharging Search with generative AI, describes it in detail. The AI snapshot gathers information from across the web for questions that require more than one source. Users can ask follow-up questions in a conversational mode. For shopping searches, it builds a product comparison overview. Ads continue to appear in their own dedicated spaces.
Google has chosen its wording carefully. It describes helping people grasp the essentials faster before exploring further, and the links in the snapshot are intentional, so that the answer directs users back to websites.
That is the stated intention. The concern is more fundamental. Many websites depend on informational searches: how long to boil a lobster, what time the waste disposal site opens, whether a permit is required to build a shed. These are precisely the questions a summary answers in two lines.
Microsoft has offered a version of this since February with the new Bing chat, which answers questions and places small numbered links below the answer. Since Bing's share of search is small, most site owners paid little attention. Google adopting the same approach is a far more significant development.
What is likely to happen to clicks?
No one outside Google knows yet, and Google describes this as an experiment for good reason. Nonetheless, several outcomes appear likely.
The feature is unlikely to remain experimental. Google does not build something of this scale to keep it on a waitlist. We expect a broader rollout to more countries and languages, possibly in a different format and under a different name.
For simple questions, fewer users will click through. Sites that mainly answer this type of question will see it in Search Console: stable impressions alongside declining clicks.
Websites will not disappear, however. Clicks will shift.
The pages most likely to be affected are thin ones: recipe filler, generic "ultimate guides" and pages written to capture a keyword with little else to offer. The pages most likely to hold their value are those where the answer is only the beginning: a booking, a map, a price list or a place a person can actually visit. A two-line summary can state that the waste disposal site opens at eight; it cannot accept a load of brush.
We recommend examining your own site now, before this reaches Canada. Which pages exist only to answer a question, and which lead to an action a person can actually take?
How can a business become a source that AI answers cite?
Each of these summaries is built from web pages, and some of those pages receive a visible link. If the format is widely adopted, that list becomes the new front page of Google, and it is considerably shorter than ten blue links.
Earning a place on that list does not require special techniques. In our assessment, it depends mainly on well-executed, conventional SEO:
- Answer the actual question clearly, near the top of the page, in the words a person would use.
- Be specific: hours, prices, service area and the actual process. Vague pages are not cited.
- Maintain sound technical foundations so the page can be crawled and understood.
- Ensure consistency everywhere. Your Google Business Profile, your website and your directory listings should all provide the same information.
We expect "AI search optimization" to be marketed as an entirely new service within the year, under a name more impressive than the underlying work. For a small business, genuine AI search optimization should primarily mean that its information is accurate, clear and easy to find.
Why do brand searches matter more when AI answers the question?
Because no summary can replace a search for your business name. Small businesses often overlook this point.
When someone types the name of a business, or "the bakery in town," they are looking for that specific business. The AI box may display its hours, but the person was already intending to visit. These are the clicks a business owns, and they result from activities that take place entirely outside Google: word of mouth, a roadside sign, a positive experience the previous summer, a newsletter or a post shared by a friend.
If generic traffic declines, additional SEO helps, and so does being a business that people remember by name. This is traditional marketing, and in a small town it already accounts for most of the work.
What content can an AI summary not reproduce?
First-hand experience. An AI summary is an average of what has already been written. If a page says the same thing as twenty others, the summary replaces it without anyone noticing.
What a summary cannot average is the business's own knowledge: photos of the actual trail, the results of testing three types of insulation in an old farmhouse, or the week when the water is clear enough to see the bottom of the lake. This is knowledge acquired by being there.
Google was already moving in this direction last December, when it added "Experience" to its quality rater guidelines. These AI answers reinforce the point. Generic content has always been weak; with a summary placed above it, such content is likely to become effectively invisible.
As a starting point, we recommend opening Search Console, identifying pages with many impressions and declining clicks, and selecting one. Add the information only your business could have written, then move on to the next page.