DerivateX analyzed 402 AI citations across 15 buyer queries. Although the two engines name many of the same tools, they cite almost entirely different sources.
A new benchmark study from DerivateX finds that ChatGPT and Google AI Overovers, two AI systems that software buyers are increasingly relying on for research, recommend many of the same products, yet cite almost entirely different sources to support those recommendations. Across open-ended purchasing questions, the two engines named the same tool about 32% of the time, but cited the same webpage only about 4% of the time.
Brands assume that wins with one AI engine will carry over to other AI engines. it’s not. ChatGPT and Google AI overviews rarely cite the same sources, so you need to gain visibility for each separately. ”
— Apoorv Sharma, Co-Founder, DerivateX
The study, titled “The Consensus Gap,” analyzed 402 individual quotes generated across 15 buyer intent queries across five segments of the cloud media and content infrastructure category. To reduce the impact of variability built into the AI’s answers, each query was submitted to both engines 5 to 10 times, and a combined set of recommendation tools and citation sources was recorded for analysis. The complete dataset is published with the report for journalists and researchers to verify and extend the findings.
Also read: AiThority interview with Matej Bukovinski, Chief Technology Officer at Nutrient
This study addresses a question that has direct commercial interest for companies seeking to be discovered through AI search: If a buyer asked ChatGPT and Google the same question, would they get the same answer? While the data partially agree on which tools are named, there is little agreement on the evidence used to justify them.
Agreement on tools, differences on evidence. For open-ended discovery questions, brand overlap between the two engines averaged 38%, but dropped to 22% for problem-solving questions. Source overlap was much lower for all question types. Even in a direct comparison query that explicitly specified the three tools, both engines were effectively pushing the same product and only shared about 30% of the cited sources. For open-ended queries, source overlap was reduced to approximately 4%.
Opposite source settings: The two engines also trust different types of pages. ChatGPT relies heavily on community discussions, with Reddit and forums accounting for about 25 percent of citations. The entire query set referenced Reddit 39 times. Google AI Overview cited only about 4% of its citations from community sources, referred to Reddit seven times, and instead prioritized vendor and competitor pages, which accounted for about 45% of its citations. Google AI Overview also cites more sources per answer, averaging about 16 compared to ChatGPT’s average of about 10, and points to the tool’s own website more frequently, with about 20 percent of citations versus ChatGPT’s 15 percent.
Use a different source universe instead of sorting on the same page. A natural explanation for the low overlap is that each engine is using the same pages but ranking them differently, with one page being treated as the primary and the other being relegated to the background. This study directly tested this and found that it did not hold. When the researchers pooled everything ChatGPT cited, both the primary citation and its supporting sources, the combined set still matched only about 25% of the sources placed front and center in the Google AI overview. Approximately 75% of Google’s primary sources were not listed anywhere in ChatGPT’s answers. This result shows that the two engines utilize completely separate evidence.
A small shared core. Despite the differences, a small number of tools consistently appeared in both engines. Across categories, Vimeo, Wistia, Gumlet, and Cloudinary were named most frequently, forming the basic set that buyers would encounter regardless of which assistant they use. However, when it comes to problem-solving questions, both engines often answer without mentioning the product name at all, suggesting that these questions are the most difficult place for brands to get mentions.
“Brands assume that wins with one AI engine will carry over to other AI engines. Data shows the opposite,” said Apoorv Sharma, co-founder of DerivateX. “ChatGPT and Google AI overviews rarely cite the same sources, which means visibility for each needs to be earned on its own terms.”
The study notes that the practical implication is that the two engines operate more like separate marketing channels rather than one AI search surface. While getting citations on ChatGPT relies heavily on your presence and sentiment in community discussions, getting citations on Google AI Overview relies heavily on structured, indexable vendor and comparison content that Google already ranks. Because the underlying evidence rarely overlaps, a single asset rarely wins in both places, and companies measuring AI visibility with only one engine may be truly misplaced. For teams evaluating external support, DerivateX also publishes a list of the best B2B SaaS SEO agencies built for the realities of these two engines.
Also read: AI Systems – Interoperable AI Systems: Connecting models across platforms
[To share your insights with us, please write to psen@itechseries.com ]
