The AI local pack names far fewer contractors than the map pack

Here is the number that should get your attention. Across one measured set of local queries, the AI local pack surfaced 5,943 unique businesses. The classic three-pack surfaced 18,330 across the same queries.

Same searches. About a third as many businesses named.

That is the whole story, and everything below is detail. The surface that increasingly answers local searches has room for far fewer companies than the surface it is displacing.

What is actually happening

AI Overviews now appear on roughly 68% of local business queries. Traditional local packs appear on about 39% of the same set. The AI answer is no longer the thing that sometimes shows up above the results. For most local searches it is the result.

For our verticals it is more pronounced still. In home services, legal and healthcare, AI Overviews intercept somewhere between 40% and 68% of searches depending on the query type. Those are the three verticals that also took the hardest ranking shifts in the March 2026 core update, which is not a coincidence. These are the categories where Google is most aggressively rebuilding how answers get delivered.

And consumers have moved with it. Around 45% of US consumers report using ChatGPT or another generative tool to get local business recommendations. That is not an early-adopter sliver anymore. That is a homeowner asking an assistant who to call about their water heater.

Why fewer businesses get named

A three-pack is a list. It has three slots, it is computed from where you are standing, and it refreshes as you move. Drive across town and a different three businesses appear. Across a metro over a day, a lot of companies get shown.

An AI answer is a recommendation. It names one company, or a few, and it explains why. It is not trying to give you options. It is trying to give you an answer.

That difference compounds. A recommendation engine has a strong incentive to name businesses it is confident about, because a bad recommendation is worse for it than a boring one. Confidence comes from corroboration: the same facts about your business, stated the same way, in several places the model can reach. Businesses that are hard to verify do not get named, because naming them is a risk.

One analysis put it at 1.2% of local business locations being recommended by AI search. Treat that specific figure with some caution, because the methodology behind numbers like this varies a lot and it is one measurement rather than a consensus. But the direction is consistent across everything we have seen: the shelf is shorter, and getting on it is a different problem from ranking.

The part that makes this worth doing now

Volume from AI search is still small next to Maps. We tell every client that, because the opposite claim is being made loudly by people selling AI SEO packages and it is not true yet.

What is true is that the leads convert better.

ChatGPT-referred inbound calls have been measured producing leads at a 49% rate, against 43% for calls originating from a Google Business Profile. AI-search referral traffic has been reported converting at roughly three times classic organic across home services datasets.

That makes sense if you think about the behavior. Someone who asked an assistant a real question, got a specific recommendation with reasons attached, and then called, is further along than someone who tapped the first pin on a map. They arrived pre-sold. Your job on that call is smaller.

So the honest framing is: small volume, high quality, growing fast, and cheap to pursue because the work overlaps almost entirely with what already makes your map pack position stronger. That is why we bundle it rather than sell it separately. Nobody should be paying a separate line item for this yet.

What actually gets a contractor named

None of this is exotic. It is the same discipline as local SEO, applied with a machine reader in mind rather than a skimming human.

Answer in the first sentence

Assistants lift sentences. They favor content that states the fact immediately and plainly. A page that opens with three paragraphs of throat-clearing before it says what the company does, where it works and what it costs, loses to a page that leads with it.

Read your own service page. If the first sentence does not say what you do and where, rewrite it. That single change is the highest-return thing on this list and it takes an afternoon.

Make the facts consistent everywhere

Your site, your Google Business Profile, your reviews and your citations need to describe one business the same way. Different phone numbers, a business name that appears three ways, a service area on the site that does not match the profile: each of these is a reason for a model to be less confident about you, and confidence is what gets you named.

This is unglamorous and it is most of the work. It is also exactly what the March 2026 core update rewarded, which tells you these are not two separate projects.

Let the crawlers in

Check your robots.txt. A surprising number of contractor sites block the AI crawlers, usually because a developer pasted in a blocklist they found somewhere. If GPTBot, ClaudeBot, PerplexityBot and Google-Extended cannot read your site, you are not in the answer and no amount of content fixes it.

Use schema that matches the page

Structured data that describes what the page actually says helps a machine parse you instead of inferring you. Schema that contradicts the page is worse than none, because it reads as unreliable.

Write the questions out

FAQ content written to be quoted rather than skimmed does real work here. Plain question, direct answer in the first sentence, specifics after. That format is close to the shape an assistant wants to lift.

What we would not do

We would not chase this at the expense of the map pack. Maps still produces the overwhelming majority of calls for every home services client we have, and a contractor who neglects their profile to pursue AI citations is trading a large certain channel for a small uncertain one.

We would not buy an AI visibility tool before fixing profile consistency. The tools report a problem you can usually diagnose for free, and the fix is the same either way.

And we would not believe anyone who tells you they can guarantee a citation in an AI answer. Nobody controls what a model says. What you can control is whether you are legible and corroborated enough to be a candidate, which is genuinely most of it.

Where this goes

The direction is not ambiguous. More local answers are being delivered as recommendations rather than lists, and recommendation surfaces name fewer businesses than list surfaces do. The contractors who are easy to verify will consolidate that shrinking shelf.

The work to be one of them is the work you should already be doing. That is the good news, and it is also why we think most of the urgency being sold around this is misplaced. Get the fundamentals right and you are a candidate. Skip them and no AI strategy saves you.