Aydın Nasuh Google AI Overviews and ChatGPT answer the same questions. They don’t read the same internet.
DataForSEO’s Search Insider 2026 report was published in October. It covers 137.9 million Google AI Overview answers and 27.5 million ChatGPT answers. That’s more than 165 million AI answers, all US and English. One part of it looks at which websites each engine cites.
The two lists barely look alike. YouTube is number one in AI Overviews. In ChatGPT it’s number 19. And more than a quarter of ChatGPT answers cite nothing at all.
I build ShubHQ, an AI visibility tool for solo SaaS founders. So I read this part of the report with one question: if you have a small SaaS and no marketing team, where should your name show up? Here’s what the data says, and what I’d do with it.
The short version
- Two different lists
- AI Overviews lean on video and social sites. ChatGPT leans on Wikipedia, reference sites and news publishers.
- One shared source
- Reddit is the second most-cited site in both.
- ChatGPT often cites nothing
- 27.3% of ChatGPT answers have no source. In AI Overviews it's 0.7%. Without sources, ChatGPT answers from what it already knows.
- What it means for you
- Being cited in one engine doesn't mean you're cited in the other. Check both, and work on the sources each one reads.
Where Google AI Overviews get their sources
These are the ten sites Google’s AI Overviews cited most:
Most-cited sites in Google AI Overviews
YouTube has more than twice the citations of Reddit. Six of the ten are video, social or community sites.
So Google’s AI likes content made by people: videos, threads, posts. It also likes its own properties.
Where ChatGPT gets its sources
ChatGPT’s list is a different world:
Most-cited sites in ChatGPT answers
Wikipedia first. Then reference sites, health sites and publishers.
One note before comparing the two charts. DataForSEO counts AI Overview citations and ChatGPT answers that cite a site. Those are different units. So compare the order, not the size of the bars.
The same site, two different ranks
Put the two lists side by side and the gaps get big:
Same site, different rank
| Site | Google AI Overviews | ChatGPT |
|---|---|---|
| youtube.com | Google#1 | ChatGPT#19 |
| reddit.com | Google#2 | ChatGPT#2 |
| wikipedia.org | Google#4 | ChatGPT#1 |
| facebook.com | Google#3 | ChatGPTnot in top 30 |
| instagram.com | Google#6 | ChatGPTnot in top 30 |
| tiktok.com | Google#10 | ChatGPTnot in top 30 |
| reuters.com | Googlenot in top 30 | ChatGPTin top 30 |
| theguardian.com | Googlenot in top 30 | ChatGPTin top 30 |
Facebook, Instagram and TikTok are in Google’s top ten. None of them is in ChatGPT’s top 30. Reuters, The Guardian, AP News and The Washington Post are in ChatGPT’s top 30. None of them is in Google’s.
The report found the same split at category level. “Internet & Telecom” is the closest category to software in the data. It’s third by citations in AI Overviews and much further down in ChatGPT. The report has no separate SaaS category, so I won’t put a SaaS number on it.
I see the same thing in my own tracking. In ShubHQ we run “best tools” questions for many categories. For “best website monitoring tools”, the LLMs named observability platforms: Datadog, New Relic, Dynatrace. Google’s AI Overview named uptime tools: Pulsetic, UptimeRobot, Better Stack. Same question, two different sets of competitors. I wrote more about that case in the SaaS positioning post.
27% of ChatGPT answers cite nothing
This is the number I’d keep from the whole section.
AI answers with no source cited
AI Overviews almost always show sources. ChatGPT skips them in more than one answer out of four.
When there’s no source, ChatGPT answers from what the model already knows. That’s training data, and it’s months old. No page you publish this week can change that answer. Only the name and description of your product that already sit across the web can.
In my last post I showed an example: a monitoring tool that was shut down in March 2026 still shows up in ChatGPT, Claude and Gemini answers. Google’s AI answers, which search the live web, had already dropped it.
What this means for a small SaaS
The report’s own advice is to look beyond your website and measure visibility on more than one surface. I agree. Here’s how I’d turn that into work for a SaaS with no team. The data part is DataForSEO’s. The plan is my view.
Start with Reddit
It's the only source near the top of both lists. Find the threads where people ask for tools like yours. Answer the question. Don't post your link unless someone asks.
For AI Overviews, make short videos
YouTube is Google's most-cited source. A two-minute video for each main use case is enough. Use your buyer's words in the title, not your feature names.
For ChatGPT, get mentioned by others
ChatGPT reads reference sites and publishers. A Wikipedia page isn't realistic for a new SaaS. Niche blogs, newsletters and "best tools" lists are.
Use one name and one sentence everywhere
For the answers with no source, ChatGPT relies on memory. Same product name, same one-line description in every directory, profile and bio.
Check both engines, separately
Ask the same buyer question in Google and in ChatGPT. Save the answer and every source it cites. Look at the two lists side by side.
Two things I’d skip:
- Writing for one engine. A page “optimised for ChatGPT” does nothing for the answers that cite no source. Those come from what other sites say about you.
- Social posts as a ChatGPT strategy. Facebook, Instagram and TikTok are big in Google’s AI and missing from ChatGPT’s top 30. Post there for people and for Google. Don’t expect them to move ChatGPT.
That screenshot is from my own project. When the engines searched the web for my category, YouTube and Reddit were the top two sources. It’s a small sample. It points the same way as 165 million answers.
Quick check for your SaaS
- You asked one buyer question in both Google and ChatGPT this month.
- You saved the sources each answer cited.
- Your product is mentioned in at least one Reddit thread you didn't start.
- You have at least one short YouTube video about your main use case.
- Your product name and description are the same in every directory.
- At least one site you don't own lists your product.
FAQ
Which sources does ChatGPT cite most?
In DataForSEO's 2026 study of 27.5 million ChatGPT answers, Wikipedia was cited most, then Reddit, Fandom, Britannica and NIH. ChatGPT leans toward reference sites and news publishers.
Which sources do Google AI Overviews cite most?
YouTube, then Reddit, Facebook, Wikipedia and Google's own pages. Six of the top ten are video, social or community sites, based on 137.9 million AI Overview answers.
Why do some ChatGPT answers have no sources?
27.3% of ChatGPT answers in the study had no citation. In those cases ChatGPT answers from its training data instead of a live web search. In Google AI Overviews only 0.7% of answers had no source.
If I'm cited in AI Overviews, will ChatGPT cite me too?
Not necessarily. The two engines favour different sources, and the study warns that visibility on one AI platform doesn't guarantee visibility on another. Check both separately.
What's the one source both engines share?
Reddit. It's the second most-cited site in both Google AI Overviews and ChatGPT.
Where I am
ShubHQ, November 2026
- LLM visibility
- 0 of 16 answers name ShubHQ, tracked since Sep 27
- Answering threads, no links to ShubHQ
- YouTube
- No videos yet
- Mentions on sites I don't own
- Directories only, 24 so far
This is the start line. I'll update the box as the numbers change.
See which sources AI uses for your category. ShubHQ asks your buyer questions in ChatGPT, Gemini, Grok and Google AI Overviews, saves every answer and lists the sites each one cited. Create a free account → Or run the free LLM visibility checker first, no signup.
Sources
- DataForSEO: Search Insider 2026 Edition (October 2026). 137.9M Google AI Overview answers, 27.5M ChatGPT answers, US, English.
- ShubHQ AI Visibility, own project data, category “SaaS Growth Tools”, Sep 27 – Nov 1, 2026
- ShubHQ Hub runs, categories “Best Website Monitoring Tools” and “Best Uptime Monitoring Tools”, Nov 2, 2026
- More from this report: 7 numbers for SaaS founders.
