The State of Local AI Search Results Near You
When people ask if I am afraid of AI Overviews killing SEO, I usually respond with something like “Local businesses are perhaps the most immune from AI Overviews; Regardless of however the AI results are derived, they have to be grounded on businesses that are actually located somewhere or the results will be even more useless than a pizza topped with rocks.”
That doesn’t mean there aren’t patterns we can observe and use. So let’s jump into the current state of AI content in Local search results.
This article covers:
- Google AI Overviews Local Search Results
- How Many Local Keywords Show AI Overviews?
- What Kind of Local Keywords Get AIOs?
- Mining Text Fragment URLs For Clues
- Common Home Services Keywords That Have AI Overviews
- Common Real Estate Keywords That Have AI Overviews
- The 3 Types of Local AI Overview Results
- Search GPT Local Search Results
- Perplexity Local Search May Be The Best of the Lot
Google AI Overviews Local Search Results
AI Overviews don’t seem to appear for many standard local queries (e.g. “restaurants near me”, “best restaurants near me”, etc.), but I thought it would be helpful to look at what Semrush reports for various locally-focus sites to get an idea of the state of these SERPs at scale. One big caveat is that Semrush’s Domain Overview reports are not really great for tracking truly local results as I think they only scrape from a small number of locations. But this data aligns with what I see from my location, so it’s likely directionally correct.
First, let’s look at how AIOs appear in keywords for some of the big local business directories:
How Many Local Keywords Show AI Overviews?
#s in 000s; Source: Semrush
| Domain | # of Keywords | # of KWs w AIOs | % of Keywords | Est Monthly Traffic | Est. Monthly Traffic - KWs w AIOs | % of Est Monthly Traffic |
|---|---|---|---|---|---|---|
| Yelp.com | 70,900 | 321 | 0.45% | 159,500 | 260 | 0.16% |
| MapQuest.com | 46,400 | 124 | 0.27% | 30,000 | 77 | 0.26% |
| YellowPages.com | 24,800 | 25 | 0.10% | 2,500 | 2 | 0.09% |
| Foursquare.com | 14,300 | 16 | 0.11% | 1,300 | 1 | 0.07% |
| Waze.com | 6,500 | 19 | 0.29% | 6,100 | 15 | 0.25% |
While Yelp and MapQuest both have >100K AIO-generating keywords, it doesn’t add up to much traffic. These sites get an estimated ~200M organic clicks/month of which <0.25% are at risk of getting hijacked by AIOs.
What Kind of Local Keywords Get AIOs?
I took 66,000 unique AIO keywords for the above domains and ngrammed them. Here are the top 10 two-word phrases the data set:
| ngram | count |
|---|---|
| zip code | 5539 |
| phone number | 1101 |
| what time | 706 |
| routing number | 650 |
| how much | 640 |
| customer service | 631 |
| area code | 471 |
| new york | 436 |
| gas prices | 357 |
| how many | 335 |
The data suggests that queries involving numbers (zip codes, phone numbers, time, gas prices, how many, etc.) are prone to showing AIOs. Zip code queries are particularly interesting as Google often will map a zip code query to a relevant city (e.g. “restaurants 94566” will often show results for “restaurants in Pleasanton” v “restaurants in 94566”). But it seems like adding the “94566 zip code” to the query will generate an AIO. I tried this for several zips and it always worked, but only for “restaurants” or “bars” queries, not for other verticals such as “realtors”, “plumbers”, etc.:
Mining Text Fragment URLs For Clues
This result provides a pretty interesting clue as to how these results are derived. If you click on the OpenTable link in the blue-shaded list on the right, you are taken to a text fragment URL that highlights the content the AI likely used to write some of the descriptions on the left-hand list. Here’s what it looks like:
If you look through the highlighted text, you can see some of the phrases used in Google’s AI Overviews for the different restaurants. Why is this interesting? Well, if you care how your business is presented in this particular result, then perhaps you should be figuring out how to influence your business’ description on OpenTable.
Let’s look at some specific verticals like Home Services. Here’s the AI Overview data for some of the top Home Services websites:
Common Home Services Keywords That Have AI Overviews
#s in 000s; Source: Semrush
| Domain | # of Keywords | # of KWs w AIOs | % of Keywords | Est Monthly Traffic | Est. Monthly Traffic - KWs w AI Overviews | % of Est Monthly Traffic |
|---|---|---|---|---|---|---|
| Angi.com | 3,400 | 199 | 5.84% | 5,400 | 275 | 5.08% |
| HomeAdvisor.com | 1,300 | 20 | 1.50% | 668 | 9 | 1.35% |
| Thumbtack.com | 859 | 10 | 1.21% | 1,400 | 5 | 0.32% |
While the numbers are still relatively small, you can see the percent of keywords and traffic are larger than they were for the general directory sites. This is likely because Home Services sites often invest in a lot of how-to content. Angi.com, in particular, is loaded with this. Let’s check out the ngrams:
| ngram | count |
|---|---|
| how much | 2838 |
| how get | 1183 |
| get rid | 684 |
| how long | 658 |
| how clean | 647 |
| how many | 577 |
| how remove | 569 |
| how get rid | 558 |
| square feet | 393 |
| average cost | 359 |
So pretty much 100% informational queries.
Let’s try Real Estate:
Common Real Estate Keywords That Have AI Overviews
#s in 000s; Source: Semrush
| Domain | # of Keywords | # of KWs w AIOs | % of Keywords | Est Monthly Traffic | Est. Monthly Traffic - KWs w AIOs | % of Est Monthly Traffic |
|---|---|---|---|---|---|---|
| Zillow.com | 13,700 | 79 | 0.58% | 84,800 | 135 | 0.16% |
| Realtor.com | 10,800 | 84 | 0.78% | 27,300 | 107 | 0.39% |
| Redfin.com | 7,800 | 56 | 0.72% | 11,200 | 79 | 0.71% |
Again, relatively small numbers.
Let’s look at the ngrams:
| ngram | count |
|---|---|
| zip code | 5601 |
| how much | 934 |
| real estate | 778 |
| area code | 731 |
| how get | 497 |
| zip codes | 373 |
| new york | 318 |
| home price | 297 |
| how many | 272 |
| get rid | 265 |
Again, these are mainly around informational queries. I don’t see anything like “single family homes for sale near me” or “open houses near me”, etc.
Instead of doing this for every vertical, let’s take a look at some of the different types of Local AI Overview results out there. There are probably plenty I have missed. DM me on LinkedIN, Twitter or Threads, if you find any more and I’ll update the post.
The 3 Types of Local AI Overview Results
There appear to be three primary formats of Local AI Overview results:
Inventory Lists:
It’s unclear how “local” this inventory is. I haven’t found any yet that appear to truly show a list of products that are clearly available in a specific city from a specific location.
Website Lists with Logos:
These seem to be mostly “national” brands and are not always businesses with physical locations.
Local Business Lists:
Here’s a variation of a Local Business Lists AIO result for a “reviews” query:
Another good reason to keep an eye on review sentiment.
And just to show how much you can trust these results, particularly when Google is not certain of your location, here’s a SERP screenshot from Semrush:
I am sure there are other variations out there, but for now, I think this gives you a pretty good idea of the current state of Google’s Local AI Overviews.
Now let’s check out:
SearchGPT Local Results
SearchGPT is OpenAI’s attempt to stab Google in the face. Let’s try the same queries we looked at for AIOs and see what we get:
National Providers for Local Results:
While SearchGPT understands my location, it doesn’t seem to have a good sense of local businesses yet, so the best it can often do is to shows results from localized pages of national sites. Note the breadcrumb link to Cars.com’s Pleasanton URL:
This query shows similar results, but this time SearchGPT maps the breadcrumb link to a relevant Google Maps URL. It’s not clear why it used Google Maps v a localized page on a relevant site like Kayak.com:
In most categories I searched, the only results/sources it showed were from national sites:
“Restaurants near me” actually showed some local business’ domains in the results. This makes sense as it’s likely restaurants, bars, etc. are probably the most searched for local biz categories in SearchGPT by several orders of magnitude. I am guessing OpenAI gives some extra weight to these sites in its algorithm, or whatever you want to call what the LLM does to determine local results.
Thus far I would say these results are far less useful than just using Google/Google Maps. I guess if you are already using ChatGPT it’s easy to click over to SearchGPT and see what comes up. It’s kind of surprising that when you do a local search in ChatGPT that it doesn’t push you over to SearchGPT or at least show the Google Maps links, but I imagine they have a lot to do over there.
Perplexity Local Search Results
Thus far Perplexity seems to have the best local search experience of all the AI services I checked:
Perplexity appears to get a lot of its local data from Yelp. It even shows Related questions under the results:
And it actually showed inventory from local dealers for this query:
That said, the inventory was likely extracted from a localized Edmunds.com page, but Perplexity was able to map it to a local dealer.
In summary, the state of Local Search Results in AI is pretty weak. Neither Google nor OpenAI appear to have good way to surface truly local business data on a consistent basis. Perplexity is kind of faking it. And I didn’t include Meta.ai for a reason. This may be why AI Overviews are so limited for various local search queries – Local Packs and Google Maps, for all their flaws do a pretty damn good job of answering the query without using LLMs. It’s not clear what added value LLM results at this point add. I imagine that will change over time as these companies figure out ways to entwine structured data with LLM data.
But for now, I think you Local SEO types should be just fine.
Share:
