Search

What local SEO actually looks like after AI overviews

The map pack did not disappear. But a growing share of local research now ends inside an answer, and the work of getting named is not the same as the work of ranking.

By Lakshay Munjal, Founder 12 May 2026 6 min read All insights →
What local SEO actually looks like after AI overviews

For fifteen years, local search had a shape everyone understood. You optimised a page, you fed the Google Business Profile, you collected reviews, and you fought for one of three slots in the map pack. The whole discipline rested on a single assumption: that a person types a query and then chooses a link.

That assumption is now wrong often enough to matter. A meaningful share of local research begins inside an AI answer — an overview at the top of the results page, a conversation with an assistant, or a question asked out loud to a phone. In those contexts there is no page two. Frequently there is no page one either. There is a paragraph, and inside that paragraph there are one or two business names.

The strategic question has changed from how do I rank to how do I get named. Those are related problems, but they are not the same problem, and the tactics diverge more than most agencies admit.

What actually changed

Three things, in order of how much they should worry you.

The click became optional. When a summary answers the question completely — opening hours, whether a clinic takes a particular insurance, roughly what a service costs — nobody visits the site. Your analytics will show impressions holding steady while sessions decline, and the natural reading is that something broke. Usually nothing broke. The answer simply got delivered upstream.

The unit of retrieval got smaller. Models don't cite pages so much as passages. A single clean paragraph that answers one question can be surfaced from a page that would never have ranked as a whole. Conversely, a beautifully optimised page that buries its answer in the ninth paragraph, after three testimonials and a founder's story, may never be quoted at all.

Consistency started outranking volume. A model assembling an answer about your business is reconciling your website, your Google profile, directory listings, review platforms and any press it can find. When those sources disagree — three different phone numbers, two spellings of the brand, an address that changed in 2023 — confidence drops, and low confidence means you get left out of the answer rather than named in it.

Ranking asks a machine to sort you. Getting named asks a machine to trust you. The second is a harder, slower and more durable thing to earn.

What we actually do about it

1. Fix the facts before touching the content

This is unglamorous and it is where most of the gain lives. One canonical name, address and phone number, expressed identically everywhere. Hours that are correct including festival closures. Services named the way customers say them, not the way your internal team says them. Every legacy listing either corrected or killed.

For a multi-location client — a hospital group, a salon chain, a builder with three site offices — this alone can take a month. It is also the difference between an assistant confidently naming you and hedging with "you may want to verify current details."

2. Restructure pages around questions

We rewrite service pages so that each meaningful question a customer asks becomes a heading, and the sentence immediately beneath it answers that question directly, in under forty words, before any context or persuasion. Persuasion still goes on the page. It just goes second.

This sounds like a copywriting preference. It is closer to an interface decision: you are formatting content for a reader who takes the first clean answer and moves on.

3. Ship structured data properly

LocalBusiness or the correct specific subtype, Organization, Service, FAQPage, and genuine Review markup where reviews genuinely exist. Structured data doesn't make you the answer, but it removes ambiguity about who you are, where you operate, and what you sell — and ambiguity is what gets you dropped.

4. Give the model something worth quoting

Generic content gets summarised generically, which means it gets attributed to nobody. Specific, checkable, original material — real pricing bands, real timelines, data from your own operations, a genuinely useful comparison nobody else has published — is what gets lifted with a name attached. This is the part clients resist most, because it means saying real numbers out loud.

5. Track prompts, not just keywords

Once a month we run a fixed set of prompts a real customer might use, across the major assistants, and log whether the client is named, cited, mentioned in passing, or absent — and which competitors show up instead. It is coarse, it moves slowly, and it is currently the only honest way to see whether any of this is working.

What local SEO actually looks like after AI overviews
What local SEO actually looks like after AI overviews

What has not changed

Almost everything else. Reviews still matter, and arguably matter more, because models read them as evidence rather than decoration. Page speed still matters. Proximity still dominates local intent — no amount of clever content makes a dental clinic in Karnal the answer for someone standing in Gurugram. And the map pack still drives an enormous share of real-world calls and directions.

Anyone selling you "AI SEO" as a replacement for the fundamentals is selling you a new invoice for old work you still need done. The correct framing is a layer on top, not a substitute.

The short version

  • Reconcile your facts everywhere before you write anything new.
  • Put the answer first on the page and the argument second.
  • Ship real structured data, not a plugin's default output.
  • Publish something specific enough to be worth citing.
  • Measure whether you get named, not only whether you rank.

How long before it shows

Listings and structured data land within weeks. Content restructuring shows in traditional search inside two to three months. Being named in AI answers is slower and less predictable — we have seen movement in six weeks and we have seen nothing for five months on accounts doing everything right. The lag is real, and it is exactly the lag most marketing budgets are not built to survive.

Which is the honest caveat to end on. This work compounds, and compounding is only visible from the far side of it.

Local SEOGEOAEOAI searchStructured data
Lakshay Munjal
Lakshay Munjal
Founder

Lakshay founded The Marketeours five years ago and still runs client delivery. He writes about search, positioning and the parts of agency work nobody puts in the deck.

Next step

Want this
done properly?

Tell us the number that has to move and we'll tell you, honestly, whether we're the right people to move it.

Start a conversation →