Asked six different ways, ChatGPT and Gemini put the same two tools at the top: Pulumi and Terraform were each named in 5 of the 6 answers we ran on 7 August 2026. The pair split first place exactly, 5 to 5, and nothing else in the category came close on frequency.
Behind the two leaders the drop is steep. AWS CloudFormation and OpenTofu were each named in 3 of the 6 answers, or 50%. AWS CDK, Crossplane and Terragrunt each landed in 2 of 6, about 33%. Everything else, including Ansible, Atlantis and Bicep, showed up in a single answer only.
One wrinkle is worth flagging before any of these counts get quoted: the assistants named some tools under two spellings. Pulumi also appeared once as "Pulumi IaC", Terraform once as "HashiCorp Terraform", and "AWS CDK" and "AWS Cloud Development Kit" are the same tool split across two labels, counted 2 of 6 each. We have kept every label as its own row rather than merge them, because merging would invent a combined number the measurement never actually produced.
The full list, counted
| Product | Named in | Share |
|---|---|---|
| Pulumi | 5 of 6 | 83% |
| Terraform | 5 of 6 | 83% |
| AWS CloudFormation | 3 of 6 | 50% |
| OpenTofu | 3 of 6 | 50% |
| AWS CDK | 2 of 6 | 33% |
| AWS Cloud Development Kit | 2 of 6 | 33% |
| Crossplane | 2 of 6 | 33% |
| Terragrunt | 2 of 6 | 33% |
| Ansible | 1 of 6 | 17% |
Six more names each appeared in a single answer only, 1 of 6: Atlantis, Bicep, InfraKit and QixDeploy, plus Pulumi IaC and HashiCorp Terraform, the two alternate spellings noted above.
How the answer changed with the question
The six questions were not synonyms, and the shortlist moved with the wording. The bare "best infrastructure as code tools" returned a tight four: Pulumi, Terraform, AWS CloudFormation and AWS Cloud Development Kit. That is the mainstream slate, and at 4 names it was one of the shortest of the six.
"What infrastructure as code tools should I use" was the widest question, returning 8 named tools in one answer: Terraform, OpenTofu, Pulumi, AWS CDK, Crossplane, Atlantis, Terragrunt and AWS CloudFormation. Atlantis appeared in this answer and nowhere else across the six questions.
"Best free infrastructure as code tools" reshuffled the deck toward open source. Its 5 names were Pulumi IaC, OpenTofu, Terragrunt, AWS CDK and Terraform. OpenTofu and Terragrunt, both genuinely free and open, rose into this set while the paid-platform framing of the leaders faded.
"Most affordable infrastructure as code tools" produced the strangest list of the six, and the most revealing. Its 4 names were Pulumi, HashiCorp Terraform, InfraKit and QixDeploy. InfraKit and QixDeploy appeared in this answer only, nowhere in the other five questions. Two names most engineers could not place surfaced purely because someone asked about price.
"Infrastructure as code tools recommendations" brought back the wider ecosystem: its 7 names included Bicep and Ansible, each of which appeared in this answer only. "Best for small teams" was the opposite extreme, narrowing to just 3 names: Terraform, Pulumi and AWS Cloud Development Kit.
Put together, five tools owed their entire presence to one specific phrasing: Atlantis to "should I use", InfraKit and QixDeploy to "most affordable", and Ansible and Bicep to "recommendations". If you measured only the bare "best" query, you would never learn that any of the five were in the running.
Where the answers came from
Across the six answers, 25 distinct source domains were cited, and one dominated. pulumi.com was referenced behind all 6 answers, 6 of 6. The second most-cited source, spacelift.io, sat behind 4 of 6, and tinyctl.dev behind 3 of 6. After that it is a long tail of roundup blogs and vendor docs cited once or twice each.
Here is the finding that matters most for anyone selling in this space: the tool cited from its own domain the most was the tool named the most. Pulumi's own site, pulumi.com, fed every one of the 6 answers, and Pulumi was named in 5 of them. It is not the only case. OpenTofu's own site, opentofu.org, fed the "best free" answer where OpenTofu was named, and in the affordability answer InfraKit and QixDeploy were each named right beside a citation to their own domains, infrakit.dev and qixdeploy.com. When a vendor's own page is in the retrieved set, its product tends to be in the answer.
What is missing from the 25 domains is as telling as what is present. No Reddit thread, no Quora thread and no Stack Overflow post appeared behind any of the 6 answers. The closest thing to community sources were github.com, cited behind 2 of 6, and a single dev.to post cited once. This category's AI answers ran on vendor documentation and roundup blogs, not forum consensus.
Review directories did not appear at all. Across all 6 answers and 25 domains, there was not one citation to G2 or Capterra. In this category the pages that decide who gets named are vendor docs like pulumi.com and developer.hashicorp.com and independent roundups like spacelift.io and tinyctl.dev, not the star-rating directories.
If you sell an IaC tool
The practical takeaway is narrow and a little uncomfortable: in this measurement the source list decided the answer, not market share. Ansible is one of the most widely used tools in the wider category, yet it was named in only 1 of 6 answers, because it sat behind far fewer of the cited pages. Getting named is about being present in the exact sources the assistant pulled.
Those sources are knowable. For these six questions they were, by reach: pulumi.com at 6 of 6, spacelift.io at 4 of 6, tinyctl.dev at 3 of 6, then developer.hashicorp.com, devopsness.com, docs.aws.amazon.com, github.com, hashicorp.com, infrasketch.cloud and sentinelone.com at 2 of 6 each. A tool written about accurately on that handful of pages has a real shot at the shortlist. A tool absent from them does not, however good it is.
QixDeploy is the whole thesis in one data point. It is not a name most engineers could produce from memory, yet it was named in the "most affordable" answer, 1 of 6, on the strength of a single citation to its own domain, qixdeploy.com. Presence in the retrieved sources, even from your own site, can put an unknown name in front of a buyer. Absence keeps a strong product invisible.
Where Glotier fits
To be straight about it: Glotier does not sell an infrastructure as code tool, and you will not find us named in any of these 6 answers. That is correct. We are not in this category, so the assistants should not be recommending us here, and they are not.
But this page is exactly what we do, pointed at someone else's category. We put six real buyer questions to ChatGPT and Gemini, recorded every tool named and every source cited, and counted. Pointed at your category, the same run shows which of the six phrasings you win, which ones you are missing, and the specific domains you would need to appear on to change that. If you want it for your own product, the check is free, no account, and takes about a minute.