Section 1
What actually broke, in numbers
Start with the evidence, because the size of the shift is the whole argument. The Pew Research Center watched the browsing of 900 US adults in March 2025. On a search results page with an AI Overview, users clicked a link 8 percent of the time. Without one, they clicked 15 percent of the time. The presence of the summary roughly halved the click. Only 1 percent of AI Overviews produced a click on a source Google had cited. People were also more likely to end their session on a page that carried a summary, 26 percent versus 16 percent. The answer satisfied them, and the traffic stopped there. The CTR studies converge on the same order of magnitude from different angles. Ahrefs, looking at the top-ranking page across a large keyword set, put the click-through drop at 58 percent when an AI Overview is present, and an earlier study measured a 34.5 percent drop in position-one CTR across 300,000 keywords. Seer Interactive's September 2025 read found organic CTR down between roughly 49 and 65 percent when the Overview appears. The exact number depends on the query mix and who is counting. The direction and the scale do not vary. When Google answers in-box, the click that used to reach you falls by roughly a third to two thirds. Now the part most operators miss. This is not staying in the "how much does a new AC cost" bucket. Semrush tracked more than ten million terms across 2025 and found the informational share of AI-Overview-triggering queries fell from 91.3 percent in January to 57.1 percent in October. Over the same window, commercial queries triggering an Overview rose from about 8 to 19 percent, transactional from about 2 to 14 percent, and navigational from under 1 percent to over 10 percent. Read that as a spreading front. The behavior that killed your informational traffic is moving down the funnel, toward the commercial and transactional queries that sit much closer to a booked job. That is the setup. The channel that was cheapest is gone, and the mechanism that took it is advancing. What follows is a way to think about it that does not collapse into either panic or the comforting consensus that you can simply "earn the citation" and be fine.
Section 2
The framework: three lenses on one collapse, and the thing none of them can see
No single model captures a regime change cleanly, so run several and state what each one misses. That discipline is the difference between a framework and a forecast you will regret. Lens one: channel-shift comparative statics The first lens is the plainest. Move one variable, hold the rest, and trace where the equilibrium lands. The variable that moved is "Google now answers the query in the results." Trace it through each channel and you get a first-order direction for each. Informational content: the click is the product, and the click is what the Overview intercepts. Direction, sharply down. Paid search and Local Services Ads: demand that used to find you free now has to be bought, so more advertisers chase the same auction, and cost per lead rises. Direction, up in cost. The map pack, meaning the three local business listings with the map: this is a result Google cannot fully answer in prose, because it resolves to "who is near me, open, and trusted," which is a ranked list of businesses, not a paragraph. Direction, roughly stable, and relatively more valuable as the surfaces around it decay. Owned channels, meaning your list, your maintenance base, your referral loop: untouched by the results page entirely, because they do not pass through a search at all. Direction, immune. This lens gets the direction of each channel right. That is its job and its limit. It tells you which way each channel moves under the shock. It does not tell you the speed, and it quietly assumes the shift settles into a new stable state. If Google keeps changing the interface, there is no new equilibrium to land on, and a static picture will lag the market. Comparative statics, the channel-shift lens • Assumes: one variable moves and the system settles into a new equilibrium you can read off. • Fits because: the AI-answer regime is a single clean shock hitting each channel differently. • Breaks when: the interface keeps moving, so there is no settled equilibrium, only a moving target. • Counteracts: treating all "digital marketing" as one exposed blob instead of channels with opposite fates. • May reinforce: a false sense that the new state is stable and can be planned around indefinitely. Lens two: does the collapse cross into your funnel (threshold and contagion) The second lens asks the question that decides how scared you should be. A collapse in informational traffic is survivable if it stays in informational traffic. The danger is contagion into the transactional queries that book work. So model it like a spread. The "infected" state is a query class that Google answers in-box instead of sending onward. The spread rate is how aggressively Overviews expand from informational into commercial and transactional intent, which the Semrush data shows is already happening. The threshold you care about is the point where the Overview footprint reaches the specific queries that put money in your account. For most service firms, the near-me and hire-intent queries are still resolving to a map pack rather than an Overview. Whitespark's May 2025 study is the sharpest data here, and it carries a nuance worth stating carefully. Across 540 local queries in Houston, Phoenix, and Denver, AI Overviews appeared on 68 percent of searches overall. That headline number is real, and it is also misleading if you stop there, because the same study found an inverse relationship by intent. On explicit local queries like "plumbers in Phoenix," Overviews showed up only 15 percent of the time while a local pack appeared 93 percent of the time. On informational queries like "how much do lawyers charge," Overviews showed up 92 percent of the time and a local pack 6 percent. The 68 percent average sits on top of that split. The queries closest to a booking are, for now, the least colonized by the Overview. That is the threshold you monitor. As long as your hire-intent queries return a map pack, the contagion has not reached your funnel and the map pack is a firebreak. The moment Google starts rendering a "here are three good plumbers near you, here is what to ask them" Overview on those queries, the firebreak is gone and the exposure is systemic. Watch that line specifically. Do not comfort yourself with the informational carnage being "only" informational, and do not assume the transactional queries are safe forever. SIR / threshold, the contagion lens • Assumes: the collapse propagates from one query class to adjacent ones at some rate, with a threshold that matters. • Fits because: Overviews are demonstrably expanding from informational into commercial and transactional intent. • Breaks when: the spread is not smooth. Google can flip a whole query class overnight by product decision, not gradual diffusion. • Counteracts: the complacency of "my booking queries still work, so I am fine." • May reinforce: over-forecasting a steady creep when the real risk is a discrete, sudden interface change. Lens three: the operators chasing the consensus (behavioral) The third lens is pointed at you and your competitors, not at Google. When a channel dies, operators herd toward whatever the loudest advice names as the replacement. Right now that advice is answer-engine optimization, the practice of structuring your content so an AI names you as a source. The pitch writes itself: if the answer lives in the box, get your business into the box. Vendors report brands seeing up to 40 percent higher visibility in AI results after adopting AEO, though that figure comes from the people selling the service and should be treated as a claim, not a finding. The behavioral trap is base-rate neglect dressed as strategy. A citation is not a customer. Pew measured it directly: only 1 percent of Overviews led to a click on the cited source. Being named in an answer the user reads and then abandons is closer to a billboard on a road nobody drives down than to a lead. AEO is worth doing as damage control, to defend the branded and reputational queries where a wrong AI answer costs you. It is not a demand channel, because it does not, by design, produce the click that a demand channel exists to produce. The herd is optimizing for presence in a surface that was built to end the session. Presence there is not the same as being chosen. Behavioral, the consensus lens • Assumes: operators systematically over-weight the vivid, socially-endorsed fix and under-weight its base rate. • Fits because: AEO is being marketed as the new SEO while its click economics are near zero. • Breaks when: an operator actually has data on their own AEO-driven conversions and can judge it on evidence rather than herd signal. • Counteracts: the pull to spend the recovery budget on the fashionable channel because everyone in the trade group is. • May reinforce: cynicism that dismisses AEO entirely, when it does have a real defensive job on branded queries. The structure-break flag Here is the thing all three lenses rest on, and the thing most commentary ignores. Every click-through model, every "at position one you get X percent of traffic" rule of thumb, every media plan built on last year's cost per acquisition, is a base rate from the pre-Overview world. The AI-answer regime voided those base rates. A ranking that returned a click in 2023 returns roughly a third to two thirds fewer today for the same position. If you are forecasting next quarter's organic leads off a 2023 CTR curve, you are planning for a market that no longer exists. When you catch a plan resting on a "normal" traffic number that used to hold, stop and flag it. That number expired. Rebuild the plan on channels whose economics did not just get rewritten.
Section 3
The solution: rank the channels, date the moves, check the history
Analysis that stops at "it is bad" is a lecture. Turn it into a decision in three passes. Pass one: rank every channel by AI-vulnerability Score each channel on one axis. How much of its value depends on Google sending a click that the Overview can intercept. The more a channel lives or dies by that click, the more exposed it is. This gives you a defund-to-defend map. The pattern is not subtle. Value is migrating from channels that depend on an intercepted click toward channels that do not pass through the results page or that resolve to something Google cannot answer in a paragraph. Your net exposure is high if your pipeline sits in the top rows and thin if it sits in the bottom rows. The reallocation writes itself once the rows are ranked: money and effort move down the table. Pass two: the dated portfolio (act without pretending to know the timing) You cannot know the month the contagion reaches your booking queries, or how fast Google pushes Overviews onto local intent. So do not stake the business on one forecast. Build a portfolio of moves sorted by how reversible they are. Do now, because they are right in every version of the future and cheap to reverse. Claim, complete, and actively work your Google Business Profile, because the map pack is the least exposed paid-free surface you have. Turn on a post-job review request so reputation compounds. Start capturing an email or phone number on every completed job so you begin building demand you own. None of these depend on a guess about Google's roadmap. All of them help whether the collapse accelerates or stalls. Zero regret. Hedge, because they are cheap insurance against the bad scenario. Convert your best repeat customers onto maintenance or service agreements now, while it is an offer and not a rescue. A recurring base that books itself is the thing that keeps the lights on if both organic and the aggregator auction turn against you in the same quarter. The cost is bounded. The protection against a demand drought is large. Defer, with a written trigger, because they are expensive or hard to reverse and you should wait for a signal. Do not tear down the site and rebuild it transactional-first today on spec. Do not sign a twelve-month AEO retainer on a vendor's 40-percent claim. Instead, pre-commit the trigger. For the rebuild, it might be "when Search Console shows my hire-intent queries losing clicks at flat impressions for two straight months." For a serious owned-media or paid push, it might be "when aggregator cost per lead crosses the point where a booked job stops clearing my margin." Write the trigger and the response now, so that when the signal fires you execute a plan instead of improvising in a panic. Pass three: the history check (what usually happens when a cheap channel dies) Base your nerve on the reference class, not the fresh wound. Marketers have watched a cheap acquisition channel die before, more than once, and the pattern rhymes. Directory and yellow-pages traffic collapsed into search in the 2000s. Facebook organic reach for business pages was throttled toward zero between roughly 2014 and 2018, which forced brands onto paid. Google's own Panda and Penguin updates wiped out thin-content traffic overnight for whole categories. In each case the same three things happened. The firms whose entire pipeline sat on the dying channel were hurt worst and some did not survive. The cost of the paid alternative rose as the displaced demand crowded in. And the firms that came through were the ones that had already built demand they owned, a list, a base, a reputation, a referral habit, before the channel died. The base rate, then, is not "the internet stops working for service firms." It is "the operators who owned a channel outlast the ones who rented one." That is the steadying fact while the organic numbers get ugly. One honest caution on the history. The matrix can break in a way the analogs do not cover. In every prior episode the click still existed somewhere. It moved from directories to search, from organic to paid, but a human still arrived at a business. The AI-answer regime is the first one that can remove the click entirely, resolving the whole transaction inside the results without anyone ever landing on a business at all. If Google moves from answering "how much does an AC cost" to brokering "book a vetted HVAC tech near you" directly, the analogs weaken, because that is a new thing and not a repeat of Facebook 2016. Watch for that move. It is the scenario your reference class has not lived through yet.
Section 4
What this framework cannot see
The honesty is the authority here, so name the blind spots plainly. This read assumes Google's Overview behavior keeps expanding along the trend the 2025 data shows. A regulatory reversal, an antitrust remedy, or a business decision to pull Overviews back on commercial queries could slow or partly undo the collapse, and the studies above are already noticing surge-then-pullback movement rather than a clean line. It assumes your local map pack stays a firebreak, which holds only until Google decides to render hire-intent queries as an Overview too. It leans on click-through studies that measure aggregates, so your own category and geography could be milder or far worse than the averages, and the only way to know is your own Search Console, not a benchmark. And it treats owned channels as immune, which is true for the results page but not for everything. A list still has to be earned and kept warm, and a referral loop still depends on doing good work. Immune from the AI-answer regime is not the same as free.
Section 5
The fitness test
You are in reasonable shape, and should play defense-plus-build, if a large share of your pipeline already comes from repeat work, referrals, and hire-intent local searches that still return a map pack, and if you can name the last five customers without checking which channel sent them. For you the collapse is a warning, not a wound. Spend the next two quarters moving down the vulnerability table and converting your base onto agreements, and you route around the whole problem. You are exposed, and should treat this as urgent, if most of your leads arrive through informational content, broad paid search, or lead aggregators, and if your organic clicks have fallen while your rankings held. That combination means the collapse has already reached you and is heading for the queries you have left. Your move is not a better AEO deck. It is to build one owned channel to the point where it books work before the aggregator auction and the transactional Overview close the last cheap door at the same time. Either way, stop asking how to win the results page. The results page is being redesigned to end the visit. The operators who come through the next two years are the ones who saw the cheapest channel die, ranked what was left by how much of it Google could intercept, and moved their money to the rows the answer box cannot reach.