Datadog was the one name ChatGPT and Gemini kept returning: it appeared in 5 of the 6 answers we measured, an 83% hit rate, more than any other data observability tool.
We ran this on 2026-08-06. We put six buyer questions about data observability software to ChatGPT and Gemini, gave each assistant one live web search per question, and recorded every product they named and every source domain behind the answers. Only those two assistants answered, so every count on this page is out of 6.
Monte Carlo came second, named in 4 of the 6 answers (67%). Below it sat a three-way tie: Bigeye, Sifflet, and Soda each appeared in 3 of the 6 answers (50%). No tool swept all six, and the space between the leader at 5 of 6 and the 3-of-6 tier is where the real story sits.
How the shortlist changed with the question
The shortlist did not hold still. It rewrote itself depending on how the buyer phrased the request, and the sharpest break came from one word: free. Asked for the best free data observability software, the assistants named just three tools, OpenObserve, Grafana, and Soda, and neither Datadog (83% overall) nor Monte Carlo (67% overall) showed up at all.
OpenObserve and Grafana are the tell. Each was named in only 1 of the 6 answers, and that lone appearance was the free question. Phrase the request any other way and both disappear. For AI buyers, "free" is close to a separate market with its own set of recommended names.
"Most affordable" produced yet another list: Dynatrace, New Relic, Datadog, and SolarWinds Observability, with SolarWinds Observability appearing in that 1 question and nowhere else. Monte Carlo, the second most named tool overall, is absent from the affordable answer entirely. Strikingly, the "best free" answer shared not one name with the "most affordable" answer. The assistants treat cheap and free as different questions, not two words for the same thing.
The bare "best data observability software" query returned six names: Monte Carlo, Sifflet, Datadog, New Relic, Dynatrace, and Splunk. Splunk turned up here in 1 of the 6 answers and in no other question, so the most generic query is the only place Splunk is mentioned at all.
"Best data observability software for small teams" named four tools: Metaplane, Monte Carlo, Bigeye, and Datadog. Metaplane appeared in 2 of the 6 answers overall, and this small-teams question was one of its two. That is the closest thing to a niche the data shows: a name that surfaces for smaller buyers but not for the generic or the free query.
The open-ended questions ran longest. "What data observability software should I use" and "data observability software recommendations" each named 8 tools, while "best free" named only 3. The looser the question, the longer the list the assistants were willing to give.
Counting across all six questions, five tools were one-question wonders: ELK stack (only in "what should I use"), Grafana and OpenObserve (only in "best free"), SolarWinds Observability (only in "most affordable"), and Splunk (only in the bare "best"). Each was named exactly once. If your product is one of these, its entire AI visibility hangs on a single phrasing.
The source map behind the answers
One domain was universal: reddit.com fed all 6 of the 6 answers. A community forum, not a vendor page or a directory, was the source ChatGPT and Gemini reached for every single time. Quora did not appear once.
Analyst and directory pages were strong but not everywhere. gartner.com was cited in 5 of the 6 answers and g2.com in 3 of the 6. Capterra never appeared at all. In this category, on this run, G2 earned its place in the source pool and Capterra did nothing.
The key insight is in the vendor domains. Monte Carlo's own site, montecarlo.ai, fed 4 of the 6 answers, and Monte Carlo the product was named in 4 of the 6. New Relic's domain, newrelic.com, also fed 4 answers, and atlan.com fed 4. When a vendor's own writing sits in the source pool, that vendor tends to land in the answer. The pages that get retrieved are the ones doing the recommending.
The reverse is the warning. groundcover.com fed 4 of the 6 answers but Groundcover was never named as a product, xurrent.com was cited in 3 of the 6 answers without its company being named, and digna.ai fed 1 answer without Digna being named. A citation is necessary but not sufficient: your content can shape the answer while a competitor takes the mention.
What we would do to get named here
If we sold a data observability tool, we would not start from market share, because market share is not what these answers are made of. They are made of a short stack of pages: Reddit (6 of 6), Gartner (5 of 6), a few vendor domains cited 4 times each, and G2 (3 of 6). Getting named is downstream of being present in those exact sources.
So the work is concrete. Show up in the Reddit threads the assistants actually pull, because Reddit was the single source in 6 of 6 answers. Publish real comparison content on your own domain, the way montecarlo.ai (4 of 6) does. Get onto Gartner and G2 and skip nothing that recurs. The tools named 5 and 4 times here are not proven to be the biggest, only the best represented in this specific source list.
The full list, counted
These are the ten tools ChatGPT and Gemini named in at least 2 of the 6 answers. Every count is taken straight from the run, out of 6.
| Product | Named in | Share |
|---|---|---|
| Datadog | 5 of 6 | 83% |
| Monte Carlo | 4 of 6 | 67% |
| Bigeye | 3 of 6 | 50% |
| Sifflet | 3 of 6 | 50% |
| Soda | 3 of 6 | 50% |
| Atlan | 2 of 6 | 33% |
| Coalesce Quality | 2 of 6 | 33% |
| Dynatrace | 2 of 6 | 33% |
| Metaplane | 2 of 6 | 33% |
| New Relic | 2 of 6 | 33% |
Glotier does not sell data observability software, so we are correctly absent from all 6 of these answers, and we would not fake our way in. What you are looking at is our measurement method run on someone else's category. It is the same run we do for a customer on their own category: the six questions, every named product, and the source page behind each answer. If you want to see where your product lands, the check is free, takes about a minute, and needs no account.