Last updated: 7 May 2026
Key statistics
| Statistic | Value | Source |
|---|---|---|
| AIO appeared in 98.1% of all runs | 309 of 315 prompt runs | Own dataset, 5 May 2026 |
| Reddit links in decision SERPs vs awareness | 178 vs 60 (about 200% more) | Own dataset, 5 May 2026 |
| External validation share grows from awareness to decision | 11.7% → 38.8% | Own dataset, 5 May 2026 |
| English prompts generate 40% more visible links during decision | 35.69 vs 25.44 | Own dataset, 5 May 2026 |
| Comparison prompts produce the broadest SERPs | 31.21 visible links on average (cluster C) | Own dataset, 5 May 2026 |
“The question is not whether you have content for every topic in the journey. The question is whether that content is picked up, cited and repeated elsewhere.”
In my article about the ChatGPT buyer journey, I used Quolity data to measure how ChatGPT actively searches for external sources before generating an answer—the so-called fan-out. That behaviour depended strongly on the buyer journey stage: during consideration, ChatGPT searched externally an average of 13.7 times, compared with just 4.7 in the awareness stage. See: The AI Buyer Journey: How ChatGPT Navigates Your B2B Brand.
I did not measure what Google was doing at the same time.
I have now done so. I analysed 315 prompt runs across four domains (GEO, HR, Supply Chain and SaaS), with three runs per prompt, and validated every related capture.json file. What I found confirms the picture from the ChatGPT analysis, but I also noticed something new.
What does the dataset look like?
The dataset contains 105 unique prompts, each run three times, for 315 prompt runs in total. For each run, I measured the number of AIO citations, organic results and all visible links on the SERP. I classified each prompt by buyer journey stage based on the intent in the wording: awareness (what is X, how does X work), consideration (who are the experts, which sources are reliable) or decision (compare the best, is X reliable, which agencies specialise in this).
This yielded 87 runs in awareness, 114 in consideration and 114 in decision.
One figure stood out immediately: Google showed an AI Overview in 98.1% of all 315 runs. Only six runs had no AIO. These were branded discovery questions in a small niche (GEO marketing), where Google apparently did not yet have enough material to create a summary. For all other prompts, AIO has become the default.
Stage 1: Awareness—does Google handle explanatory questions differently from ChatGPT?
In the awareness stage, Google generates an average of 11.31 AIO citations and 25.93 visible links per run. AIO presence is 96.6%.
The substance of those citations is at least as interesting as their number. In the awareness stage, 38.2% of AIO sources are encyclopaedic or explanatory content—the highest share of all three stages. The rest consists mainly of vendor homepages and introductory articles. The external validation share—the proportion of sources outside the vendor itself—is lowest here: 11.7% of AIO citations.
Google therefore treats awareness prompts as discovery questions. The AIO explains, and the organic results offer more detail. Reddit is among the five most visible domains, but with 60 link appearances it is still relatively modest.
This aligns with ChatGPT’s behaviour at this stage: less fan-out (an average of 4.7 external searches), more explanatory answers. Both platforms act like well-informed assistants explaining a concept without immediately looking for confirmation.
What this means for a B2B brand: visibility during awareness requires explanatory content that is clear, well structured and easy to find. Vendor content still works well here. But if you are visible only at this stage, you are missing the rest of the journey.
Stage 2: Consideration—who gets to join the conversation?
The consideration stage has the highest AIO presence of the three stages: 99.1%. That is no coincidence. An AIO appears almost every time prompts ask who the experts are, which sources are reliable or who writes about a topic.
Average citations increase to 13.31, and visible links to 26.52. But the shift in source mix is more striking than the increase. The share of external validation in AIO citations rises from 11.7% to 26.3%. Trade media, consultant directories, LinkedIn profiles and niche publications replace vendor explainers. In the top five AIO citations at this stage are consultancy.nl (cited 38 times) and Frankwatching (37 times).
Here, Google behaves less like a search engine and more like an authority map: who is cited, who writes and who advises? A brand visible only on its own website hardly stands out at this stage. The AIO cites what others say about you, not what you say about yourself.
ChatGPT does something similar during consideration, but more actively: a 100% fan-out rate, with an average of 13.7 external searches. That is the highest of all stages in the ChatGPT dataset. Both platforms actively look for external confirmation of authority—Google through the SERP structure, ChatGPT through fan-out.
What this means for a B2B brand: look beyond SEO on your own domain. If your brand does not appear in trade media articles, consultant directories or LinkedIn discussions, an AI platform is unlikely to mention it in the consideration stage. Content alone is not enough; its distribution matters too.
Stage 3: Decision—when does external validation take over?
Almost everything reaches its maximum in the decision stage: an average of 14.26 AIO citations, 28.68 visible links and an external validation share of 38.8% in organic results.
The biggest surprise in the data was the explosion of Reddit. The visible SERPs contained 60 Reddit links in awareness, 98 in consideration and 178 in decision. Almost three times as many. Google’s signal is clear: for decision prompts, forum validation, price comparison sites and peer reviews rise to the surface.
The top citation domains in the decision stage are reddit.com (61 citations), vendor-specific sources (slimstock.com, 47; linkedin.com, 35) and comparison platforms. The AIO therefore cites vendors, but also the places where buyers actually talk: forums, reviews and shortlist pages.
The same pattern appears at run level. The prompt “Is Miro a reliable collaboration platform for enterprise teams?” produced a SERP with Miro itself, G2 reviews and Reddit threads—three different kinds of confirmation side by side. A brand cannot fully control any of the three.
ChatGPT shows slightly less fan-out in the decision stage than during consideration (an average of 9.48 searches), but the questions are more concrete and comparison-focused. Both platforms do the same thing: gather sources that do not come from the brand itself as external confirmation.
What this means for a B2B brand: if you are absent from G2, Gartner, Capterra or relevant Reddit discussions, you are missing a significant part of the decision SERP. This is not a matter of writing one more article; it is a structural question of reputation beyond your own content stream.
How does language affect the breadth of visible SERPs?
One unexpected detail: English prompts consistently generate broader SERPs than Dutch ones.
In the decision stage, English prompts produced 35.69 visible links compared with 25.44 for Dutch prompts—a 40% difference. During consideration, the gap was a little smaller but still present (33.86 versus 24.86).
A plausible explanation is that the English-language publishing landscape is larger and deeper, with more comparison sites, forums and trade publications. This means a B2B brand active internationally or publishing English-language content faces a richer and more competitive SERP—and has more to prove before it is cited.
Does the query cluster predict SERP structure better than the industry?
One hypothesis I wanted to test was whether prompts from different industries (HR, GEO, Supply Chain and SaaS) behave differently in the SERP. The answer: less than the query clusters do.
Comparison prompts (cluster C) produced the broadest SERPs, with an average of 31.21 visible links. Branded discovery questions (cluster A) were the narrowest, with 22.79 visible links and the lowest AIO presence (93.7%). Expert and thought leadership questions (cluster E) consistently attracted expert and consultant profiles, regardless of whether the topic was HR or SaaS.
This is a useful insight for marketers deciding what content to create. The way a question is phrased partly determines the SERP structure. A comparison question activates different signals from a discovery question, even when both concern the same topic.
What remains for a B2B brand?
The common thread I see in the ChatGPT analysis and now the Google AIO data is the same: the further a prospect moves through the buyer journey, the less Google and ChatGPT rely on what you publish as a brand, and the more they draw on what others say about you.
During awareness, vendor content still works well as a foundation. During consideration, trade media and your visibility as an expert partly determine whether you are cited at all. During decision, forum validation, review platforms and comparison sites become a structural hurdle—not something you can get around with a few more blog posts.
Another striking finding: brand mentions in AI Overviews were zero per cent in the cross-domain decision data. Not low—zero. Direct brand mentions in AIOs appeared to some extent only for branded discovery questions in the GEO domain itself. As soon as the prompt became more general or more decision-oriented, the direct brand mention disappeared.
The implication is simple, but requires a structural approach: a GEO strategy that stops at your own website is incomplete. The question is not whether you have content for every topic in the journey; the question is whether that content is picked up, cited and repeated elsewhere.
Frequently asked questions
At which stage of the buyer journey do Google AI Overviews cite the most external sources?
The external validation share is highest in the decision stage: 38.8% of organic results fall outside the vendor sphere. During awareness, that is only 11.7%.
Is AIO presence really as universal as the data suggests?
Based on 315 prompt runs, yes. An AI Overview appeared in 98.1% of runs. It was occasionally absent only for branded niche questions in a very small market (GEO marketing). For most B2B topics, you can assume AIO is standard.
How do ChatGPT and Google AIO treat the B2B buyer journey differently?
ChatGPT actively uses fan-out: it searches for external sources before answering. Google selects more passively, building an AIO from what is already indexable. But the pattern by journey stage is strikingly similar: more external validation as a decision approaches.
Why is Reddit so visible in the decision stage?
For decision questions, users look for peer validation: “Has anyone actually used this, and what was their experience?” Reddit answers exactly that question. Google recognises the signal and includes Reddit in AIO and organic results.
Does it matter whether you write content in Dutch or English?
Yes. English prompts consistently generate broader SERPs. During the decision stage, there are 40% more visible links. If you want international visibility, an English-language content strategy is hard to avoid.
What does the cluster analysis tell us about content strategy?
Comparison content (cluster C) produces the broadest SERPs—more than industry or domain type. How a searcher phrases a question determines the SERP pattern that follows. Understanding which content types fit which query clusters helps you anticipate the structure Google will show.
What is the most concrete action to take based on this data?
Map which providers are already cited in AIOs for decision prompts in your market. Those are the places—trade media, directories and forums—where you should also be present. Not through ads, but with useful contributions that are picked up externally.
Sources
- Own dataset: 315 Google AI Overviews prompt runs collected on 5 May 2026, across GEO, HR, Supply Chain and SaaS.
- Quolity data on ChatGPT fan-out behaviour: The AI Buyer Journey: How ChatGPT Navigates Your B2B Brand.
- Each capture.json per run contains aiOverviewPresent, citations[], organic[] and allVisibleLinks[]. All summary.csv values were validated against their corresponding capture.json files; 0 mismatches found.
By Hans Schepers—a digital marketing professional specialising in GEO (Generative Engine Optimisation) and AI visibility for B2B brands.
This translation was made by AI.