Section 1
Three markets, three starting structures
Before the models, get the facts of the divergence straight, because the whole argument rests on them. The United States got the collapse first and fullest. AI Overviews rolled out broadly to US users in 2024, and AI Mode, the more aggressive conversational version, went through Search Labs in March 2025 and reached general US searchers by mid-June. US service firms were mostly standing on two channels when it hit: Google organic search, and lead aggregators like Angi and Thumbtack. Both are directly exposed. The informational organic died in the box, and as displaced demand crowded the aggregators, cost per lead rose. The United Kingdom got the AI product almost as fast, because it is an English-language market outside the EU's regulatory perimeter. Google launched AI Mode in the UK on 29 July 2025, and notably skipped the Search Labs gating it had used in the US, pushing it straight to searchers. So on timing, the UK broke much more like the US than like the EU. But the UK sits on a different demand structure. British trades run heavily through paid directory platforms, Checkatrade, Rated People, MyBuilder, where members pay in the region of 60 to 120 pounds a month plus per-lead fees, and where a single lead can be sold to several tradespeople at once. The UK operator was leaning on a paid directory more than on free organic content, which changes what the AI shock actually threatened. The European Union got the collapse late and in pieces. AI Overviews reached EU users roughly nine to ten months after the US, appearing in only eight member states plus Switzerland in March 2025, with large markets including France initially excluded. Google itself has said the EU's rules, the Digital Markets Act, the Digital Services Act, and the AI Act, hold back the launch of new features. So EU service firms have been living through a slower, staggered, partial version of the same shock, protected less by anything they did and more by the friction their regulator imposed on Google. Three markets. The US hit early on an organic-plus-aggregator base. The UK hit early on a directory base. The EU hit late on friction. Now run the models.
Section 2
The framework: one shock, three structures, and the regulator nobody can model
Use three lenses and say what each cannot see. The mistake to avoid is treating the US timeline as a law of physics that Europe will simply repeat on a delay. Lens one: same shock, different equilibria (comparative statics across a cross-section) The first lens is comparative statics run sideways. Instead of moving one variable through one market, hold the shock constant and move the starting structure. The same "Google answers in-box" lands differently depending on what channel the firm was standing on. For the US firm on organic-plus-aggregator, the shock hit both exposed channels at once: free organic clicks fell, and the aggregator auction got more expensive as everyone crowded in. That is the doom loop in its purest form. For the UK firm on a paid directory, the direct hit is smaller and the indirect hit is real. The informational-organic collapse matters less because the UK trade was already renting leads from Checkatrade rather than earning them from free content. But the same crowding pressure that raises US aggregator prices raises UK directory prices, and members already report renewal hikes, with some accounts describing jumps from around 50 pounds to 150 pounds a month. The UK operator was insulated from the organic collapse by a directory that is now itself getting more expensive for the same reason. Different channel, same squeeze arriving by a different road. For the EU firm, the shock is the US shock with a lag. The structure is broadly the same Google-dependence as the US, but the regulator delayed and fragmented the rollout, so the equilibrium is arriving in slow motion and unevenly across member states. This lens gets the direction of each market right. Its limit is that it treats each structure as fixed. It assumes the UK stays directory-led and the EU stays delayed, when both can change, and the thing most likely to change them is the regulator, which this lens does not contain. Comparative statics, the cross-section lens • Assumes: each market's channel structure is fixed while the shock passes through it. • Fits because: US, UK, and EU firms demonstrably stood on different channel mixes when AI search arrived. • Breaks when: the structures themselves shift, especially if regulation rewires how Google may present results. • Counteracts: the lazy import of US tactics into markets with a different channel base. • May reinforce: a static regional map that misses the structures converging over time. Lens two: the EU delay is a lag, not immunity (threshold and contagion) The second lens models the spread and asks what the EU delay actually buys. It is tempting for a European operator to read "AI Overviews came late and partial to the EU" as protection. Model it as a contagion and that reading falls apart. The DMA and related rules act as friction, lowering the effective rate at which the AI-answer behavior spreads across the EU. Friction slows a spread. It does not confer immunity. The threshold, the point where the collapse reaches the queries that book jobs, is deferred, not cancelled. The EU is early in the curve the US is already deep into, and the base case is convergence, because Google's incentive to unify its product across markets does not disappear. The UK is the tell here. Being outside the EU's regulatory perimeter, it crossed early, on nearly the US timeline. That is the natural experiment: strip away the DMA friction and an English-language market gets the collapse fast. The EU's slower curve is the DMA holding the rate down, not the disease being milder. A European firm treating the delay as a reprieve is misreading a lag as a cure. SIR / threshold, the contagion-lag lens • Assumes: the collapse propagates across markets at a rate regulation can slow but not stop. • Fits because: the UK's near-US timing outside the DMA shows the delay is regulatory friction, not milder demand. • Breaks when: regulation does more than slow diffusion and actually blocks or reshapes the product, which is a different mechanism than a lag. • Counteracts: the false comfort that the EU is structurally protected rather than merely behind. • May reinforce: an assumption of smooth convergence when the catch-up could arrive as a sudden jump rather than a creep. Lens three: false comfort and imported playbooks (behavioral) The third lens points at the operators. Two predictable errors show up in Europe. The first is status-quo bias dressed as prudence: the UK trade leaning on its Checkatrade listing and the EU trade enjoying the rollout delay, both concluding that the thing everyone is panicking about in America is not really their problem. The delay and the directory feel like moats. They are buffers with a clock on them. The second error is the opposite and just as costly: importing the US consensus wholesale, in particular the answer-engine-optimization gold rush, into a market where it maps badly. A UK firm whose demand runs through Checkatrade does not primarily have an AI-citation problem. It has a directory-dependence problem, and spending its recovery budget on getting cited by ChatGPT solves the wrong bottleneck. The discipline is to diagnose your own structure before adopting anyone's playbook. The US advice is written for a firm that stood on organic and aggregators. If you stood on a directory, or if your regulator has frozen the shock in place for another year, your first move is different. Behavioral, the imported-consensus lens • Assumes: operators over-weight both the comforting local buffer and the loud foreign fix, and under-weight their own structure. • Fits because: the delay and the directory invite complacency while US-authored AEO advice invites misallocation. • Breaks when: an operator actually measures their own channel mix and conversion and acts on that evidence. • Counteracts: copying US tactics without checking whether the US channel structure is yours. • May reinforce: analysis paralysis, using "my market is different" as a reason to do nothing. The structure-break flag Here is what none of the three lenses can hold, and what the imported US commentary ignores entirely: in Europe, the regulator is an active player who can rewrite the game, and that voids any straight-line extrapolation from the US timeline. This is not a background condition. In the UK, the Digital Markets, Competition and Consumers Act came into force on 1 January 2025, and on 10 October 2025 the CMA designated Google with strategic market status in general search and search advertising. AI Overviews and AI Mode were explicitly named as within scope. By early 2026 the CMA had proposed conduct requirements including fair treatment of publisher content in AI-generated responses and search choice screens. In the EU, on 9 December 2025 the European Commission opened an antitrust probe into Google's AI search tools, with publishers arguing the Overviews use their content while demoting them. Either regime could force Google to alter how Overviews work in a way that has no US analog at all. The US structure-break was Google's product decision. The European structure-break could be a regulator's order. You cannot model the second by watching the first.
Section 3
The solution: diagnose the structure, date the moves, check the history
The levers, by market The reallocation differs by where you stood when the shock hit. For the US firm, the levers are the ones in the core AI-immunity playbook: defund informational content built for clicks, treat the aggregator as a throttle rather than a base, and invest first in the map pack and owned demand. For the UK firm, the primary lever is directory de-dependence, not AI citation. The Checkatrade or Rated People listing is a paid channel that is getting more expensive for the same crowding reason US aggregators are. The move is to build the free, resilient channel underneath it: claim and work the Google Business Profile, because the local pack survives the AI shock for the same proximity-and-prominence reasons everywhere, and to convert directory-sourced customers into a repeat and referral base you do not pay per lead for. AEO is a low priority for a UK trade whose bottleneck is directory rent, not answer-box visibility. For the EU firm, the primary lever is time. The rollout delay is a genuine, if temporary, window that US and UK firms did not get. The correct use of a window is to build the owned channels, the list, the reviews, the map-pack prominence, before the catch-up arrives, rather than to assume the window is a wall. The EU operator who spends 2026 building owned demand is using the DMA's friction for exactly what it is worth. The one who treats it as safety will get the US collapse late and unprepared. The dated portfolio Do now, in every market, zero regret. Claim and complete the Google Business Profile and start a steady review request. The local pack is the least-exposed surface in all three regions, and this work is correct whether the collapse is early, late, or frozen. For the UK firm specifically, add: start capturing the customer's direct contact on every directory-sourced job, so you stop re-renting the same customer. Hedge, cheap insurance. Keep the directory or aggregator running at a capped spend as a bridge while owned demand builds, in the UK and US alike. Bounded cost against a pipeline gap. For the EU firm, the hedge is to not switch off any working channel on the assumption the delay lasts. Defer, with a trigger. Do not rebuild a whole site or sign a long AEO retainer on the US timeline. Pre-commit the trigger to your own market's signal. For an EU firm: "when AI Mode reaches full rollout in my country and my Search Console shows hire-intent clicks falling, then I execute the transactional rebuild." For a UK firm: "when my directory renewal crosses the point where a booked job stops clearing margin, then I cut directory spend and shift it to owned." Write the trigger to your regulator's reality, not to America's. The history check The base rate for Google feature rollouts across regions is well established, and it is the honest reference class. Google ships features US-first, the EU lags by months to about a year, sometimes longer under regulatory pressure, and then the markets converge. Regulation has repeatedly slowed European adoption of a Google product. It has rarely reversed it. The base rate says the EU delay is measured in quarters, not permanence, and that the UK, being outside the EU regime, converges to the US faster. Plan on convergence with a lag, not on divergence forever. The caution, and it is the live one, is that the matrix can break in Europe in a way the base rate has never covered. The DMCC conduct requirements and the DMA antitrust action could force Google to run a materially different search product in the UK or EU: real choice screens, enforced fair treatment of publisher content in AI answers, data portability. If that happens, the European search market stops being "the US market on a delay" and becomes its own structure with no US analog. That is the scenario your reference class cannot price, and it is not far-fetched given a live CMA designation and a live Commission probe. Watch the regulators as closely as you watch Google, because in Europe they are the other author of the outcome.
Section 4
What this framework cannot see
Name the blind spots. This read assumes the regulatory actions stay roughly on their announced course, when a change of government, an appeal, or a negotiated settlement could soften or harden them abruptly, and the CMA's conduct requirements were still at the proposal stage in early 2026. It assumes the UK's directory culture holds, when British trades could migrate toward Google and owned channels faster than expected, changing the structure mid-analysis. It leans on rollout dates and on click studies drawn largely from US and English-language data, so the behavior of AI Overviews in German, French, or Polish search may differ in ways the English-language studies do not capture. And it treats the EU as one market for simplicity, when the member-by-member rollout means a firm in one of the eight early countries is living a different timeline than a firm in a country still excluded.
Section 5
The fitness test
You are a UK firm that should act now, and treat the directory as the problem rather than AI as the problem, if most of your work comes through Checkatrade, Rated People, or MyBuilder and your renewal cost has been climbing. Your bottleneck is directory rent, not answer-box citation. Spend the next two quarters building the map pack and an owned base so you can throttle the directory instead of depending on it. You are an EU firm that should use the window if AI Mode has not yet fully rolled out in your country and your organic clicks are still roughly intact. You have a lead that US and UK firms did not get. Use it to build owned demand and map-pack prominence before the catch-up, and pre-commit the trigger that tells you the window has closed. You are exposed and should move fastest if you are a US firm, or a UK firm on organic-plus-aggregator, whose informational clicks have already fallen while rankings held. The collapse has reached you, the regulatory counterweight forming in Europe does not protect you, and your move 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 together. Either way, stop reading the US collapse as a universal timetable. The shock is the same. The structure it hits, the channel you stood on and the regulator standing over Google, is not, and in Europe that difference is large enough to change what you do first.