Every AI engine has a personality. And the brands winning AI visibility have figured out which one they’re talking to.
I used to think of “AI search” as one channel. Our data changed my mind.
When we looked at where six AI surfaces pull their sources from, the differences were big enough to change how I’d build a strategy from scratch.
The split
At Foundation we track how ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews and Google AI Mode cite sources across B2B software categories. Across roughly 380,000 answers between June 22 and July 24, 2026, here’s what stood out:
- Google’s search-embedded surfaces, AI Overviews and AI Mode, cited community and social sources 3 to 5 times more than the standalone assistants.
- Perplexity and Claude were the most review-heavy surfaces, with review sites making up about 21% of their citations.
- Across every surface, product and company websites took the biggest slice, between 50% and 71% of citations.
Here’s the split in one table.
| Source type | Google AI Overview and AI Mode | ChatGPT, Claude, Gemini and Perplexity |
|---|---|---|
| In 39% to 52% of citing answers | Claude cited it zero times | |
| YouTube | In 38% to 47% of citing answers | Claude cited it zero times. It barely registers on ChatGPT and Gemini |
| Social domains | About 7% of citation share | 0.3% or below |
| Review platforms | Under 10% of citations | 16% to 21% of citations on ChatGPT, Perplexity and Claude |
| Vendor and product pages | Still the largest source on every surface | 52% to 72% of citations |
So your own site still matters most. But the remaining slice, the third-party sources, looks completely different depending on who’s answering.
Why Google leans on the crowd
Google’s AI surfaces sit on top of Google Search. And Google Search has spent years pushing Reddit threads, forum posts and YouTube videos higher in results.
So when Google’s AI builds an answer, it pulls from the same well. Real people talking. Threads. Videos. Community discussion.
That’s why Reddit kept a strong position in Google AI Mode and AI Overviews even after ChatGPT pulled back on it in August.
Why Claude and Perplexity lean on reviews
The standalone assistants behave differently. Claude and Perplexity reach for structured, third-party evaluation. G2. Capterra. TrustRadius.
On Claude, review platforms were the only external category with meaningful reach in our data. If your brand isn’t on the review sites, Claude has very little outside evidence to work with.
What Google’s rankings add
Citations are one half of the Google story. Rankings are the other. In September we pulled Ahrefs data on 300 “best [X] software” searches where Reddit ranks in the top 20 and each term gets 200 or more searches a month. Together those searches add up to 496,950 a month in the US.
| Reddit result | Keywords | Share of 300 |
|---|---|---|
| In the top 3 | 196 | 65% |
| Top 3 as a blue link | 73 | 24% |
| Top 3 as an AI Overview or module citation | 123 | 41% |
| Not in the top 3 | 104 | 35% |
Look at how Reddit wins. On 24% of those searches it holds a blue link in the top three. On 41% it shows up inside the AI answer. Reddit holds position 1 on 119 of the 300 keywords, and 113 of those, or 95%, are AI Overview citations.
B2B looks different. Here is how Reddit’s footprint on B2B keywords moved over six months.
| Reddit on B2B keywords | March 2026 | September 2026 |
|---|---|---|
| Top-3 count across 191 B2B keywords | 131 | 107 (down 18%) |
| Finance and accounting top-3 rate (40 keywords) | 68% | 38% |
| “best small business accounting software” (63,000 searches a month) | #1 | #4 |
Reddit looked like it improved on 53 B2B keywords. On 40 of them, about three in four, the gain was a citation inside the AI Overview. The other 13 climbed the blue links.
Reddit is slipping in B2B rankings while Google keeps quoting it inside AI answers. Citations and rankings tell different stories, and your plan needs both. Rankings show whether Google still trusts Reddit pages. Citations show whether the AI answer quotes them. This is one snapshot of 300 keywords, so treat it as a direction. It is a sample of 300 keywords, and the market is bigger. ChatGPT’s change was far sharper, and I cover it in ChatGPT Cut Its Reddit Citations in Half.
Your category changes the picture too
The surface is only half the story. The category matters just as much.
In our data, CI/CD and managed database topics leaned toward community sources. Developers trust other developers, and those conversations live on Reddit and in forums.
Recruiting software, SEO platforms and employee engagement tools leaned toward review sites. The spread between topics was about 3.3x for reviews and about 2.5x for community sources.
And the two effects stack. When we drilled into recruiting software, the review lean of the category and the review lean of a surface like Perplexity combined. That’s why one brand can be highly visible on one surface and almost invisible on another.
Three companies, three different plans
Here’s how this plays out in practice. These are illustrations, built from the patterns in our data.
1. A developer tools company
The category leans community. Developers ask each other for recommendations on Reddit and in forums. If their buyers use Google AI Mode, the community lean gets even stronger. This company should put serious effort into Reddit, technical YouTube content and open documentation that answers questions directly.
2. An HR tech company
Recruiting software leans hard on reviews. If their buyers use Perplexity or Claude, review sites become the main external source. This company should build a steady review engine on G2 and Capterra and keep its comparison pages sharp.
3. An SEO platform
Another review-heavy category, with buyers who often use several AI tools. This company needs strong review coverage for Claude and Perplexity plus community and video presence for Google’s surfaces. It’s the most expensive mix of the three, because it has to win in both worlds.
A simple map for where to show up
Here’s how I’d think about it if I were building an AI visibility plan today.
1. Start with your own site
Product and company sites earn the majority of citations everywhere. Clear product pages, honest comparison pages and content that answers the questions buyers ask are still the foundation.
2. Map your category
Figure out whether your category leans community or reviews. We’ve built a lookup across 81 B2B categories for exactly this reason.
3. Match the surface your buyers use
If your buyers live in Google, invest in Reddit, YouTube and community presence. If they use Claude or Perplexity, put the work into review sites first.
4. Cover both if you can
Most B2B buyers use more than one AI tool. The brands with the strongest visibility show up in both worlds.
How to find out which AI your buyers use
Most companies guess. You don’t have to.
1. Ask in sales calls
Add one question to discovery: “Did you use any AI tools while researching options?” Then ask which ones. After a month of calls, you’ll have a real picture.
2. Add AI to your attribution form
If you have a “how did you hear about us” field, add ChatGPT, Perplexity, Claude and Google AI as options. Self-reported attribution is imperfect, but it beats guessing.
3. Check your referral traffic
Look for visits from chatgpt.com, perplexity.ai, claude.ai and gemini.google.com in your analytics. Google’s AI surfaces are harder to isolate, so treat this as a partial view.
Measure by surface
Once you know where your buyers are, measure each surface separately. A single “AI visibility score” hides the differences that matter most. Your brand might be thriving in Perplexity and invisible in AI Mode, and an average would tell you everything is fine.
Where this is heading
I expect these personalities to keep shifting. August already showed us how fast a single surface can change its sources. ChatGPT cut its Reddit citations by more than half in about two weeks, while Google kept citing Reddit heavily.
So the plan has to be built for movement. Track each surface monthly. Watch for changes in which source types each one favours. When one surface shifts, move budget toward the sources it now prefers.
The brands that win this decade will treat AI visibility like a portfolio. Several bets, measured separately, rebalanced often. That takes more work than picking one channel and going all in. It’s also the only approach that held up this year.
The trade-off
This makes AI visibility harder. One playbook won’t cover every surface. You need to spread your effort, and that means more channels, more coordination and more patience.
I’d still take that over the alternative. Betting on one surface and hoping it never changes is how a lot of Reddit-only strategies took a hit this August.
This is the distribution layer of AI search in action. Create once, then go show up wherever your buyer’s AI is listening.
FAQ
Which AI engines cite Reddit the most?
Google AI Mode and AI Overview. Reddit appears in 39% to 52% of their citing answers. On ChatGPT, the share of answers citing Reddit fell from 21.41% in July to 10.05% in August. Claude cited Reddit zero times in our data.
Why do Claude and Perplexity lean on review sites?
Review platforms made up about 21% of citations on both. I can describe the pattern. I can’t tell you the reason, because the engines don’t publish how they pick sources.
Does my own website still matter?
Yes. Product and company sites took the biggest slice on every surface, between 50% and 71% of citations. The third-party slice is where the surfaces differ.
How do I know which AI my buyers use?
Ask in sales calls, add AI tools to your attribution form and check referral traffic from chatgpt.com, perplexity.ai, claude.ai and gemini.google.com. The steps are above.
Should I optimize for all six surfaces?
Start with your category and your buyers. Then cover both worlds if you can. The 81-category lookup shows where your category leans.
How we got the data
Foundation tracks AI citations through AirOps across six AI surfaces and 81 B2B software topics. Figures come from roughly 380,000 answers collected June 22 to July 24, 2026. Domain categories (product sites, reviews, community and social, educational, media) were assigned per cited domain. Category spreads compare the most and least review-heavy and community-heavy topics in our representative set. Disclosure: Foundation has a commercial relationship with AirOps.
The ranking section uses a separate Ahrefs snapshot of 300 “best [X] software” keywords in the US, September 26 vs. March 26, 2026. It is a different dataset from the AirOps citation data.
Related reading
- ChatGPT Cut Its Reddit Citations in Half. Here’s What Our Data Shows.
- G2 Is the Quiet Kingmaker of AI Search
- Which AI Engine Should Your Category Care About? What 81 B2B Software Categories Reveal
- AI Visibility Statistics 2027
- The GEO Playbook
- The Reddit B2B Marketing Blueprint
Data comes from Foundation’s research with AirOps and Ahrefs. Citation data window: June 22 to July 24, 2026.