Google Search Is Turning Into AI Mode: What the Latest Changes Mean for Your Business

If you have looked at your Google traffic reports in the past week or two and felt that something is off, you are not imagining things. Over the summer Google shipped two quiet changes that together reshape how people see search results and how businesses measure them. AI Overviews now expand on their own, without anyone clicking anything, and the links in search results no longer point where they appear to point. Both changes push ordinary search closer to Google AI Mode, the chat-style version of search that Google has been promoting all year.

Neither change arrived with a keynote or a blog post. One was spotted by an SEO consultant on X and confirmed by Google a day later. The other went through two months of quiet testing before Google admitted it was real. Still, for business owners these two updates matter more than most of the loud AI launches this autumn, because they touch the two things every website depends on: how much traffic search actually sends you, and whether you can trust the numbers that tell you.

In this post we unpack both changes, look at what they break, and go through what a business should do about it in practical terms. Let's sink in!

What changed in Google Search this summer

The first change concerns AI Overviews, the AI-written answer block that sits on top of many search results. Until late August, Google showed only a short snippet of that answer, and you had to press a "Show more" button to read the rest. Around 27 August users started noticing that the button was gone. For many queries the full AI answer now unfolds by itself, and an "Ask anything" follow-up box loads right under it, inviting you to continue the conversation instead of scrolling to the links below.

Google confirmed the change the next day. The official wording was careful: "For some queries, AI Overviews may dynamically expand for topics where our systems determine it's most useful for people." In plain language, Google decides when you get a full AI answer, and on those queries the classic organic results move a long way down the page.

The second change is easier to miss. Since late June, Google has been wrapping the links in its search results in redirects that run through google.com/goto. The address you see under a result is still the real one, but the link you actually click is an encoded token that only Google can read. By the end of August the rollout covered close to one hundred percent of searches, and Google confirmed it officially on 26 August. Add the news that AI Mode itself got link carousels for breaking stories and a faster Gemini model behind it, and the direction becomes hard to ignore.

Why Google AI Mode is becoming the default

For over a year Google kept AI Mode as a separate tab, something curious users could try and everyone else could skip. That period seems to be over. Removing the "Show more" button, auto-expanding the answers and planting a follow-up box in the middle of the results page all do the same thing: they make the standard search page behave like Google AI Mode without asking the user to switch.

Google's Robby Stein explained that the expansion only happens where their systems find it useful, and that it stops if you have already started scrolling. That sounds considerate, but it does not change the substance. The default experience of searching Google is turning into reading an AI answer, with regular websites appearing as supporting citations rather than the main event.

Barry Schwartz of Search Engine Roundtable, who has covered Google for two decades, put it bluntly: this is not good for publishers, not good at all. For an ordinary business the effect is subtler but real. Informational queries, the ones that used to bring steady blog traffic, will produce fewer clicks. Ranking first among the blue links means less when the blue links start below the fold. What matters more and more is whether the AI answer cites you, because those few citation slots are the new first page.

The /goto redirects, or why your rank tracker looks broken

The goto change deserves a closer look, because it quietly breaks a whole layer of tools that businesses rely on. Rank trackers, SEO platforms and SERP APIs have always worked by reading the search results page and parsing out which URLs rank where. With the new redirects there is nothing to parse. The clickable link is an opaque token, and finding the destination means following every single redirect, one request at a time.

The numbers involved are not small. Checking one keyword down to position one hundred used to take around ten requests. With redirect resolution it takes somewhere between five hundred and a thousand. One large SERP data provider reported that it can still resolve nearly all classic organic links, but more than half of the links inside AI Overviews and a quarter of local pack links remain unresolved. Some smaller tracking tools have simply started returning empty results.

There is also a legal backdrop. Google is currently in litigation with SerpAPI, one of the biggest search scraping providers, and the goto rollout is widely read as part of the same effort to lock down programmatic access to results. Whatever the motive, the practical outcome for businesses is the same: rank tracking is about to get more expensive, less frequent, or both. If your SEO reports suddenly show wild ranking drops this month, check with your tool vendor before you blame your website.

What this means for your analytics and attribution

The redirects raise a second, less discussed risk. When a visitor reaches your site through google.com/goto, the referrer information the browser passes along can change. If the redirect strips those signals, analytics platforms like GA4 may start counting some organic visitors as direct traffic. As of early September nobody has confirmed systematic undercounting, but the mechanism is now in place, and it costs nothing to protect yourself.

The good news is that Google Search Console is unaffected. It measures impressions and clicks on Google's side of the fence, before any redirect happens. That makes it the most trustworthy source of organic data you have right now.

The sensible move is to establish a baseline this week. Note the current ratio between clicks in Search Console and organic sessions in GA4, and check it weekly. If the two start drifting apart, or if google.com/goto shows up as a referral source in your reports, you will know the redirects are bleeding into your attribution and can adjust your reporting before anyone makes a bad decision based on broken numbers.

How to keep your visibility while search changes

None of this means SEO is finished. Commercial queries still show ads, shopping results, local packs and plain links, and people who are ready to buy still click through. What changes is the role of informational content. A blog post used to be a traffic engine; now it is increasingly a citation source, something the AI answer quotes and links when a potential customer asks a question in your field.

Earning those citations rewards the same things that always made content good: original information, clear structure, actual expertise, pages that are current rather than recycled. Generic summary articles were already struggling, and an auto-expanding AI answer removes their last reason to exist. A page that contains a number, a case, or an opinion that exists nowhere else still gets picked up.

It also makes sense to spread your measurement and your channels. On the measurement side, lean on Search Console, server logs and brand search volume rather than any single tool. On the channel side, this is one more argument for the boring, reliable assets: an email list, a community, direct client relationships. Traffic that arrives without Google's permission is worth more every month.

What teams building digital products should take from this

If you are building a digital product, there are a few concrete lessons in this story. First, any product that depends on scraping Google, whether for rank tracking, market research or price monitoring, just saw its unit economics change by two orders of magnitude. If that describes part of your stack, model the new request costs now and look at official data sources before your supplier passes the increase on to you.

Second, if your marketing site converts blog readers into signups, start measuring signups and leads rather than sessions. Traffic will look worse before conversions do, and teams that only watch sessions tend to cut the content that was quietly producing customers.

Third, build measurement that does not depend on the referrer header. Server-side tracking, tidy UTM discipline on every campaign, and conversion tracking tied to accounts rather than visits will survive whatever Google does to its links next. We build these foundations into client projects by default now, and the last two months show why.

Final notes

Google is not hiding where search is going. The results page is becoming an AI assistant that reads the web on the user's behalf, and two small interface changes this summer did more to advance that than any product announcement. Links are still there, but they are becoming the footnotes.

The businesses that will handle this well are the ones that treat it as a measurement problem and a positioning problem, not a catastrophe. Baseline your numbers this week, ask your tool vendors the hard questions, and keep publishing the kind of content an AI answer has to cite. And if you would like help adapting your site, your analytics or your SEO strategy to search as it now works, that is exactly the kind of work we do at Davydov Consulting every day.