We Measured Where AI Assistants Get Their Answers About Local Business Software

When someone asks ChatGPT or Google about business software, which sites does the answer cite? We pulled third-party data on six search phrases to look.

We had a selfish reason to look: we wanted to know whether our own site was among those sources, and what would change it if not. It wasn't among the top cited sources in any of the lists. We're publishing the numbers so you can check them.

The short version: YouTube was the most-cited source in every Google list, and Reddit made every list on both platforms. Company websites — the thing many businesses spend their AEO and SEO budget on — made the lists too, mostly the sites of established phone and software companies. Ours didn't make any.

What the Data Measures, Exactly

We chose six search phrases for the kind of software a local business buys:

• AI receptionist for small business
• AI answering service
• virtual receptionist service
• AI appointment scheduling
• missed call text back service
• AI chat agent for website

On 30 June 2026 we ran each phrase through DataForSEO's LLM Mentions data, once for ChatGPT and once for Google's AI Overviews. DataForSEO is a data company that keeps its own collection of AI responses. For each phrase, it finds the responses in that collection where the phrase's words turn up — in the question, the answer, or the searches the assistant ran behind it — and counts how often each source site was cited in those responses.

That matters for how to read everything below. These aren't answers to six questions we typed in. They're totals across many responses that matched the words. DataForSEO doesn't say when those responses were collected; 30 June is when we pulled the totals. It also returns only the most-cited sites and most-named brands for each list, not every one. We kept the top eight sites from each list, plus the brands named in the ChatGPT responses. Google's side returned no brand names in our pull.

Six phrases, two platforms, one pull of data: twelve lists. That's a small sample from one industry on two of the major assistants. It shows a pattern; it doesn't prove a law. A few other limits matter. The ChatGPT data covers US English only. The counts are DataForSEO's tallies, and Google's run far higher than ChatGPT's, so compare sources within a platform, not across. Because the lists stop at the top, they show which sites were cited most; they can't show that a site was never cited. And word matching lets in noise: a medical research archive ranks third on Google for "AI appointment scheduling," which looks like an overlap of words rather than a real buying source. We left it in.

The Headline: Two Platforms, Two Diets

Here are the three most-cited sources for each phrase, with DataForSEO's count in brackets:

Two sites showed up again and again, and neither is a company website. Both are places where people talk to each other or show each other how things work. But the two platforms' lists looked quite different.

Google's AI Overviews cited video most

YouTube was Google's most-cited source for all six phrases, usually by a wide margin. Reddit was in Google's top three for five of the six. After that came a mix that included phone-system companies, software sites and job boards.

Google owns YouTube, so in hindsight this may not be a surprise. But it changes what "doing content" should mean if Google's AI Overviews matter to you.

ChatGPT's lists were mostly text

YouTube didn't appear in ChatGPT's top eight for any of the six phrases. Reddit appeared in all six — first or tied for first in four, second in the other two. The rest of ChatGPT's lists leaned on reference sites and the tech press: Wikipedia showed up for three phrases, Forbes for three and TechRadar for four.

That's a diet of community discussion, reference material and the press. Almost all of it is text.

Company websites made the lists too

Company sites weren't absent. Business phone companies did well on Google: Nextiva was in Google's top eight for four of the six phrases, and for "AI receptionist for small business" Nextiva and RingCentral came second and third, behind only YouTube. On ChatGPT, the top source for that same phrase was getnextphone.com, an AI receptionist company's own site.

So a company website can get cited. In these lists, the ones near the top were mostly established phone and software companies, plus that one AI receptionist seller on ChatGPT. A small local business is up against them, not just against other small businesses.

What This Actually Means If You Run a Local Business

The instinct when someone says "we need to show up in AI" is to write more pages. This data suggests that instinct is incomplete.

Here's the uncomfortable part. We're a company that sells AI automation. We have a technically clean website — structured data, fast pages, every AI crawler explicitly welcomed, a real blog. We also have real clients. And yet, in the 30 June 2026 data, our site wasn't among the top cited sources in any of the twelve lists, and Athena Automation wasn't among the brands the ChatGPT lists returned. That doesn't prove we were never cited, because the lists only show the top. It does show we weren't there.

A well-built website is table stakes. It's what lets a search engine understand what you do. On this data, it isn't enough to put you among the sources AI responses cite most. In these lists, the top spots mostly went to sites run by other people — video, community discussion, reference sites, the press and job boards — not to the companies selling the product.

That points to four practical shifts:

1. Community presence counts

Reddit made every list on both platforms. Being genuinely present in the communities where your customers ask questions may do more here than another article on your own domain. Genuine is the operative word — most communities will remove you quickly for showing up to advertise, and they're right to.

2. Video matters for Google

YouTube was the most-cited source in Google's lists for every phrase we checked. A business with no video is absent from the source those AI Overviews responses cited most. For a local service business this doesn't mean production budgets. It means answering the questions customers actually ask, on camera, findable.

3. Get written about where assistants cite

ChatGPT's lists leaned on Wikipedia, Forbes, TechRadar and similar publications. A fair mention in a review roundup or a trade article is a place AI can cite, and you don't have to own it. It's slower to earn than a page on your own site, but it's the kind of source these lists favored.

4. Watch your wording

"Virtual receptionist service" pulled in job boards. Indeed was the top source on ChatGPT and second on Google, and FlexJobs and ZipRecruiter made both lists too. That many job-board citations suggests a good share of the matching responses were about finding receptionist work, not buying the service. Before you build content around a phrase, check whether job seekers or another audience use the same words. If they do, the AI responses that share your phrase may be answering their question rather than your buyer's.

A Note on Which Brands Got Named

We also looked at which brands were named in the matching ChatGPT responses. Google's side returned no brand names in our pull. The counts were small — no brand appeared more than six times for any phrase — and these lists stop at the top too, so treat this as a sketch.

Where a phrase described a product clearly, the brands fit it. "AI chat agent for website" returned Tidio, Intercom, LiveChat, Tawk.to, Zendesk and Drift, and "AI receptionist for small business" returned Dialpad, RingCentral, Smith.ai and Twilio, among others. "AI answering service" returned OpenAI, Anthropic, Google and Concentrix.

"Missed call text back service" returned a bank, a motel chain and an investment firm. That's a reminder of how the data works, not a finding about assistants: it collects responses where the words turn up, whatever the conversation was about. Those names tell you about the word match, not about what an assistant recommends for missed-call software.

Frequently Asked Questions

Where do AI assistants get their answers about business software?

In DataForSEO data for six software search phrases, pulled on 30 June 2026, YouTube was the most-cited source in Google AI Overviews responses for all six, and Reddit was among the top cited sources in every list on both ChatGPT and Google. ChatGPT's lists also leaned on Wikipedia and tech publications such as Forbes and TechRadar. Company websites appeared too, mostly those of established business phone and software companies. The data counts sources cited in AI responses that matched the phrases' words; it is not a set of answers to questions we asked.

Does publishing blog posts get you cited by AI?

This data can't answer that directly. It shows that the most-cited sources were YouTube and Reddit, with reference sites, the press, job boards and a handful of established vendors also on the lists. Your own pages help a search engine understand you, but much of what these lists cited was published by someone other than the company selling the product.

What is the difference between SEO and answer engine optimization?

SEO is largely a first-party game — your pages, your structure, your keywords. Answer engine optimization is mostly a third-party game: whether the places an assistant cites already mention you. A business can have a well-built website and still not be among the sources AI responses cite most. That was our own position in the 30 June 2026 data: our site wasn't among the top cited sources in any of the twelve lists.

How did you measure AI citations?

On 30 June 2026 we pulled DataForSEO's LLM Mentions data for six search phrases, separately for ChatGPT and Google's AI Overviews. DataForSEO matches each phrase's words against its own collection of AI responses and counts how often each source site was cited in the matching responses. For ChatGPT it also lists the brands those responses named. It returns only the top sites and brands for each list, and it doesn't state when the responses were collected. That made twelve lists, and we kept the top eight sources from each.

Is this sample big enough to draw conclusions from?

No. Six search phrases, two assistants and one pull of data can show a pattern worth acting on, but they can't settle the question. A test worth trusting needs more assistants, more questions and repeated pulls over time.

The Bottom Line

If you want AI assistants to cite or recommend your business, this data suggests a lot of the work isn't on your website. It's on the platforms their responses already cite — community discussion, video, review sites, and coverage by people who aren't you.

That is slower and less comfortable than publishing another page. It is also where the data points. We published our own position in it because a scoreboard nobody can check isn't worth much.

Correction: An earlier version of this article reported percentages from weekly captures in July 2026 and said company websites barely registered. We couldn't trace those figures to retained records, so this version uses the 30 June 2026 data, which shows company sites did appear, and describes how that data was gathered.