For a decade, SEO for local service businesses followed a predictable script: win the map pack, rank a handful of money pages, build citations and links, repeat. Google’s expansion of AI Overviews — and the parallel rise of answer engines like ChatGPT, Perplexity, and Gemini — has quietly broken that script. The query still gets typed. The intent is still high. But increasingly, the first thing a searcher reads is a synthesized answer that may never send a click to anyone.
Service verticals are where this shift bites hardest, because the stakes per query are enormous. Consider legal search, one of the most competitive and expensive corners of the web. A single high-intent phrase can be worth hundreds of dollars per click in paid search, which tells you exactly how much a provider is willing to pay for the organic equivalent. That economic pressure makes legal a useful stress test for what AI search is doing to every trust-sensitive industry — home services, healthcare, financial advice, and beyond. If you want to see the mechanics at the page level, this lawyer SEO guide lays out how the fundamentals are changing, and the patterns it describes generalize well past law.
So what actually changes when an AI Overview sits on top of the results?
Clicks compress, but qualified clicks concentrate
The reflexive panic is “AI Overviews will kill our traffic.” The reality is more nuanced. Informational, top-of-funnel queries — “what is a contingency fee,” “how long do I have to file a claim” — are exactly the kind of questions an Overview answers completely. Those clicks were never going to convert at a high rate anyway. What the Overview really does is strip away the low-value impressions that used to pad your traffic reports and flatter your dashboards.
The queries that still drive a click are the ones where the searcher needs to evaluate a provider, see proof, or take an action the model can’t take for them. That is good news if your content strategy was built around buyers, and bad news only if you were counting on informational volume to justify the whole program. The work now is to ensure your brand is the entity the model cites in its summary and the destination it recommends when the user is finally ready to choose.
Being citation-worthy is the new ranking
AI Overviews are extractive. They pull from sources the system already trusts and that express information in clean, attributable, factual units. Four things consistently make content extractable:
- Direct, declarative answers near the top of a section. Lead with the answer, then explain. Models reward the sentence that resolves the query, not the 200-word windup preceding it.
- Entities, not just keywords. Name the practice areas, jurisdictions, statutes, procedures, and people explicitly, and connect them. The model is building a graph; give it clean nodes and edges.
- Verifiable specifics. Numbers, dates, definitions, and ordered steps get lifted into answers far more often than vague reassurance.
- Structural clarity. Descriptive headings phrased the way people ask questions, plus schema where it genuinely applies, make a passage cheap for a machine to parse and quote.
This is where E-E-A-T and AI search collide. The experience and expertise signals Google has pushed for years — author credentials, original analysis, first-hand detail — are the same signals that make a passage safe for a model to quote. You are no longer optimizing for a position on a page; you are optimizing to be the trusted sentence inside the answer.
Vertical case study: how AI Overviews behave in legal SERPs
Watch a basket of legal queries over a few weeks, and a pattern emerges. Procedural and definitional questions trigger Overviews almost universally, and they cite a mix of large publishers, government sources, bar association sources, and the occasional firm blog that explains a concept unusually clearly. Commercial queries — “best [practice] attorney in [city]” — show Overviews far less consistently, and when they do appear, they lean on review aggregators and the local pack rather than long prose. A closer breakdown of which query types surface answers, and how firms are restructuring content to earn citations, is documented in this analysis of AI Overviews for lawyers.
The transferable lesson: segment your keyword universe by whether an answer engine can fully satisfy the intent. For “can be fully answered” queries, your goal is citation and brand recall — measured in impressions and assisted conversions, not raw sessions. For “must evaluate a provider” queries, you still fight for the click, and the click now lands on a page that has to do more persuasive work than ever, because the visitor arrives later in their decision and is far better informed.
It also pays to audit how you already appear. Run your priority queries and record whether an Overview fires, who it cites, and whether your brand is mentioned at all. That baseline turns a vague anxiety into a punch list: queries where you’re cited and should defend the position, queries where a competitor is cited and you could plausibly displace them, and queries where the Overview leans on sources you can realistically join. Most teams discover they’re closer to citation than they assumed — usually one or two structural fixes and a stronger proof section away from being the passage the model chooses to quote.
Don’t forget the engines that aren’t Google
AI Overviews dominate the conversation because Google still owns most of the search market, but ChatGPT and Perplexity are now meaningful discovery surfaces for exactly the research-heavy, “help me understand my options” questions that precede a high-value purchase. These systems reward many of the same things — clarity, specificity, and recognizable entities — but they also weigh brand mentions across the open web more heavily than classic link equity. Showing up consistently in industry publications, directories, and credible third-party coverage now does double duty: it builds traditional authority and teaches search engines that your brand belongs in the consideration set.
What to actually do this quarter
- Re-segment your keywords by answerability. Tag each cluster as informational-answerable, commercial-evaluative, or transactional, and stop reporting them as one undifferentiated traffic number.
- Restructure your best informational pages for extraction. Answer-first sections, question-shaped headings, and tight factual sentences, a model can lift verbatim.
- Invest in entity clarity. Consistent NAP, Organization, and Person schema that ties authors to credentials, and internal links that reinforce which topics you own.
- Strengthen the proof layer on evaluative pages. Results, reviews, original data, and specificity about who you serve and where — the content an Overview can’t replace, and a buyer can’t skip.
- Change your measurement. Track citations and branded-query lift alongside clicks. A page that sheds sessions but starts appearing in Overviews and feeds branded search is winning, not losing.
Where the prize actually moved
AI Overviews don’t eliminate SEO; they relocate the prize. The reward used to be a ranking and the click that came with it. Now the reward is being the source the machine trusts and the brand the buyer remembers when the answer engine hands them off. For low-stakes queries, that’s a clean trade — you give up clicks you couldn’t monetize for visibility you can. For high-stakes, high-value service searches, the click survives, but only the brands that built genuine authority and a persuasive destination will capture it. The providers treating this as an existential threat are mostly the ones who never had real authority to begin with. The ones treating it as a forcing function — to be clearer, more specific, and more genuinely expert than everyone around them — are quietly consolidating the visibility their competitors are panicking about losing.
