Datadog, Dynatrace and Elastic APM each landed in 5 of the 6 answers we collected when we put buyer questions about application performance monitoring tools to ChatGPT and Gemini, the highest score any product reached. No single product was named in all six.
We ran six real buyer questions on 3 August 2026, gave ChatGPT and Gemini one live web search each, and wrote down every product named and every source domain cited. Across the 6 answers, 23 product names appeared, though two of them, Amazon CloudWatch and CloudWatch, are the same AWS product recorded under two labels. The top three aside, the field thinned quickly.
Below the leaders, New Relic and Uptrace were each named in 4 of 6 answers, and AppDynamics and Grafana in 3 of 6. Everything else landed in 2 answers or fewer: SigNoz in 2 of 6, and 15 more products in a single answer each. So two-thirds of the named products owe their entire presence to just one question out of the six.
The question you ask rewrites the shortlist
The most useful thing we saw is that changing one word in the question almost completely changes the shortlist. The bare "best application performance monitoring tools" question returned eight commercial names: New Relic, Datadog, Dynatrace, AppDynamics, Elastic APM, Amazon CloudWatch, Dotcom-Monitor and Grafana. It reads like the enterprise incumbents' list.
Ask instead for the "best free application performance monitoring tools" and 6 of those 8 names drop out. That answer named Jaeger, Zipkin, Uptrace, Grafana, Tempo, Loki, Prometheus and Elastic APM, an almost entirely open-source stack. Only Grafana and Elastic APM carried over from the bare "best" answer. The four biggest commercial names, Datadog and Dynatrace at 5 of 6 overall plus New Relic and AppDynamics, vanished completely from the free question.
The "most affordable" question produced a third list again: Datadog, Uptrace, CloudWatch, SigNoz, Chronosphere, Dynatrace, IBM Instana and Quickwit. Here the two commercial leaders Datadog and Dynatrace returned, sitting beside cheaper and newer names like SigNoz, Chronosphere, IBM Instana and Quickwit. "Free" and "affordable" are not the same query to an assistant: the two 8-name lists share only Uptrace between them.
Fifteen products earned their whole presence in this data from a single one of the six questions. The free question alone is responsible for five of them: Jaeger, Zipkin, Tempo, Loki and Prometheus appear nowhere else. The affordable question alone introduced Chronosphere, CloudWatch, IBM Instana and Quickwit. "What should I use" was the only place Apteligent and Bugsnag surfaced, the small-teams question was the only home for StatusCake, the recommendations question the only one for Splunk APM, and the bare "best" question the only one for Amazon CloudWatch and Dotcom-Monitor. SigNoz was the rare middle case, named in 2 of 6, and both were the small-teams and affordable framings.
Where the answers came from
Three domains fed every single answer. blazemeter.com, newrelic.com and reddit.com were each cited in 6 of 6, so the assistants leaned on one load-testing vendor's blog, one APM vendor's own site, and one community forum in all six responses.
Community discussion sat behind every answer we collected. Reddit was cited in 6 of 6 answers and Quora in 1 of 6, the latter in the free question. A stack of roundup and blog pages did the rest of the work: pantheon.io in 4 of 6, cloudzero.com in 3 of 6, and catchpoint.com, dotcom-monitor.com and twilio.com in 2 of 6 each. These "best APM tools" listicles are where most of the named products actually come from.
The big software review directories barely registered. G2 was cited in only 2 of 6 answers, Gartner in 1 of 6, and Capterra did not appear at all. If you assumed your G2 or Capterra profile was doing the work of getting you named by an AI in this category, the data says it mostly is not.
The key insight is in the vendor domains. Five named products also had their own websites feeding the answers: newrelic.com (cited in 6 answers), uptrace.dev (4), dynatrace.com (3), dotcom-monitor.com (2) and datadoghq.com (1). Uptrace is the cleanest case in the whole run: uptrace.dev is cited in exactly the four questions where the Uptrace product is named, and in neither of the two where it is not. A small vendor reached 4 of 6 answers largely because its own domain was one of the sources the assistants read.
New Relic's content team pulled the same lever at scale. newrelic.com was cited in all 6 answers, including the free and affordable questions where the New Relic product itself was not named. Their pages are being read as a source even when their product is not the pick, which is the exact position a content strategy is built to reach.
What this means if you sell an APM tool
The takeaway from 6 answers is blunt: being present in these specific sources is what gets you named, not your market share. Uptrace was named in 4 of 6 answers, one more than AppDynamics at 3 of 6, despite being a much smaller and newer project. The difference is that uptrace.dev and the roundups it sits in were in front of the assistants.
So the work is concrete. Get your product into the roundup pages these answers actually used: blazemeter.com, pantheon.io, cloudzero.com and catchpoint.com. Show up in the Reddit threads that were cited in all 6 answers. Publish the comparison content on your own domain that New Relic and Uptrace are getting cited for. The source list, not the size of the company, decided who got named in this run.
The full list, counted
| Product | Named in | Share of 6 answers |
|---|---|---|
| Datadog | 5 of 6 | 83% |
| Dynatrace | 5 of 6 | 83% |
| Elastic APM | 5 of 6 | 83% |
| New Relic | 4 of 6 | 67% |
| Uptrace | 4 of 6 | 67% |
| AppDynamics | 3 of 6 | 50% |
| Grafana | 3 of 6 | 50% |
| SigNoz | 2 of 6 | 33% |
| Amazon CloudWatch | 1 of 6 | 17% |
| CloudWatch | 1 of 6 | 17% |
We kept Amazon CloudWatch and CloudWatch as two separate rows on purpose. They are the same AWS product, but the assistants wrote it two different ways in two different questions, 1 answer each, and we will not merge them into a "2 of 6" the measurement never actually produced.
Glotier is not on this list, and that is correct
We do not sell an application performance monitoring tool, so Glotier is rightly absent from all 6 answers. An assistant naming us for "best APM tools" would be wrong, and every product above earned its place in a category we do not compete in.
The run itself is the point. This is exactly what we do for the category you do sell in: put the real buyer questions to ChatGPT and Gemini, record who gets named and which sources fed each answer, and show you the gap between the roundups that decide the list and the ones you are actually in. If you want to see your own category measured like this, the check is free, no account, and takes about a minute.