• AI visibility

Guide · AI search

How do DevOps companies get recommended when engineers ask AI which tool to use?

By being easy for AI assistants and coding agents to find, read and trust: complete public documentation, working examples on GitHub, honest comparisons, and a reputation in the communities engineers already consult. DevOps tools are usually adopted by engineers first and bought by the enterprise later, so an AI answer that names you can start a free trial that becomes a platform contract. Engineers also distrust AI output more than most buyers do, which makes verifiable technical proof the deciding factor.

By Underneath Team Guide Updated

The short version

  1. Engineers drive the purchase: 48% of developers endorsed or influenced a new technology purchase in the past year, in Stack Overflow’s 2025 survey (opens in a new tab) of more than 49,000 developers.
  2. They already work through AI: 84% use or plan to use AI tools in development, and ChatGPT (82%) is the most used assistant, Stack Overflow reports (opens in a new tab). Yet 46% distrust the accuracy of AI output.
  3. Platform buying is now the norm: 90% of organizations have adopted at least one internal platform, according to Google Cloud’s 2025 DORA report (opens in a new tab) of nearly 5,000 technology professionals.
  4. Small teams become large contracts: GitLab had 1,456 customers paying more than $100,000 a year and a dollar-based net retention rate of 118% at the end of fiscal 2026.
  5. AI already sends signups to developer platforms: Vercel (opens in a new tab) said in 2025 that ChatGPT was bringing in around 10% of its new signups, against 1% six months before.

Who picks a DevOps tool, who pays for it, and how large can the account grow?

Engineers choose it, platform teams standardize it, and procurement signs; a won account often grows for years.

The buying journey usually runs bottom-up. In the Stack Overflow survey, nearly half of developers endorsed or influenced a technology purchase last year, and one in five of those influenced a substantial addition to their company’s tech stack. The tools they pick (continuous integration, deployment, infrastructure as code, observability, incident response) then meet a platform team that decides what the whole company will standardize on. Security tools travel a similar path with a second buyer; see how AppSec vendors win demand.

That platform layer is now common. DORA found 90% of organizations have adopted at least one platform, and that a high-quality internal platform is tied to an organization’s ability to unlock the value of AI. The CNCF and SlashData (opens in a new tab) found 28% of organizations have a dedicated platform engineering team. Separately, the CNCF reports (opens in a new tab) that the share of developers working without formalized DevOps or platform practices fell from 20% to 12%, and that the cloud native developer community reached 19.9 million.

GitLab’s results show how team adoption turns into enterprise revenue. At the end of fiscal 2026, GitLab passed $1 billion in annual recurring revenue and reported:

Customer sizeCustomersGrowth over the year
More than $5,000 of annual recurring revenue10,6828%
More than $100,0001,45618%
More than $1 million15526%

The largest accounts grew fastest, and a dollar-based net retention rate of 118% means existing customers spent more each year. For a DevOps vendor, the first team that adopts the tool is the start of a long expansion, not the end of a sale.

Where in an engineer’s day do AI assistants already show up?

Everywhere in the daily workflow, and increasingly in documentation reading, where agents now read more than people do.

  • Assistants are standard. Stack Overflow found 51% of professional developers use AI tools daily. DORA found 90% of technology professionals use AI at work. The assistants themselves compete for those engineers, as how AI coding tools win developers explains.
  • AI is a learning channel. 44% of developers used AI tools to learn to code, up from 37%. Technical documentation remained the top learning resource, at 68%.
  • Agents read docs. Mintlify, which hosts documentation sites, reports on its live data page (opens in a new tab) that agent readership grew 7.7 times from February to August 2026, while human readership grew 1.2 times. Mintlify sells documentation tools, so treat this as vendor data.
  • Signups already arrive from ChatGPT. At Vercel, a deployment platform, ChatGPT referred around 10% of new signups, up from 4.8% the previous month, the company reported.

DevOps work is also where engineers are most cautious about AI. Stack Overflow found 76% of developers do not plan to use AI for deployment and monitoring, the most resisted task in the survey. We infer that engineers will use AI to shortlist tools but verify the answer against documentation, peers and a trial before trusting a tool with production.

Which questions do engineers ask AI about DevOps tools?

Questions about fit, migration, alternatives, standards, pricing and security. We wrote the example prompts below to show how a platform engineer or SRE might phrase these questions; they are not logged from real users.

StageIllustrative prompt
Category“What is the best CI/CD setup for a monorepo with 200 engineers on Kubernetes?”
Alternatives“What are the main alternatives to Jenkins for a team moving to the cloud?”
Comparison“GitHub Actions or GitLab CI for a company that self-hosts its runners?”
Standards“Which observability tools accept OpenTelemetry data without vendor-specific agents?”
Migration“How hard is it to move our Terraform state to another infrastructure-as-code tool?”
Pricing“Which incident management tools charge per responder rather than per seat?”
Compliance“Which deployment platforms have SOC 2 Type II and support FedRAMP environments?”

A coding agent may ask similar questions on a developer’s behalf while it works, reading docs directly, as Mintlify’s data suggest. For API vendors, how agents choose an integration covers that case in depth.

Standards questions matter more than they used to. The CNCF announced in May 2026 that OpenTelemetry had graduated, describing it as allowing organizations to change analysis tools without rewriting code (opens in a new tab). When switching is easier, the shortlist is reopened more often.

How does a tool named in an AI answer end up as an enterprise DevOps contract?

Through a product-led path: answer, documentation, free use, team adoption, platform standard, enterprise contract.

  1. An engineer asks an assistant, or an agent searches, for a tool to solve a specific problem.
  2. The answer names a few tools and links documentation, GitHub repositories or community threads.
  3. The engineer reads the docs and tries the free tier or open source version.
  4. The team adopts it; the platform team evaluates it for company-wide use.
  5. Procurement and security review the vendor, and an enterprise contract follows, then expansion.

The AI answer matters most at steps 1 to 3, where no salesperson is involved. Vercel’s numbers show that step can convert at scale. The value lands much later: GitLab’s 118% net retention shows how accounts grow once a tool becomes part of the platform. That gap in time makes attribution hard; we cover it in how to prove GEO caused sales.

Why does an assistant recommend one CI/CD or observability tool over another?

The assistants don’t say; research points to independent and community sources, and engineers reward proof they can run themselves.

What the search providers publish. According to Google, AI Overviews and AI Mode may use a “query fan-out” technique (opens in a new tab), so one question about replacing Jenkins can spawn multiple related searches. OpenAI writes that ChatGPT search typically rewrites a question (opens in a new tab) into targeted queries for its search partners. Neither company explains how a particular pipeline or monitoring tool gets picked.

What studies have found. Independent sources dominate: on the US software questions examined by Chen and colleagues (opens in a new tab), AI search drew 72.7% of its sources from earned sites, against 45.4% for Google. Engineering forums count too: our Reddit study found 35.4% of Google AI Overviews on B2B software searches cited a Reddit thread, and specialist communities supplied 58.7% of the Reddit threads shown in software searches. And ChatGPT went hunting for reviews or ratings in 46.2% of its answers in our hidden-searches study.

What engineers trust. These factors are specific to the field:

  • Documentation they can run. Documentation is the top learning resource for 68% of developers. Complete, current, example-rich docs are what both an engineer and an agent read before choosing.
  • Peer recommendation. In the CNCF radar, 91% of developers said they would recommend GitHub Actions to peers. Engineers repeat what peers endorse in forums, talks and repositories.
  • Open standards and portability. Support for standards such as OpenTelemetry signals that a buyer is not locked in.
  • Skepticism of claims. 46% of developers distrust AI accuracy and 75% said not trusting AI’s answers is the top reason they would still ask a person for help. Marketing claims that cannot be tested are discounted twice.

Our inference. A DevOps vendor’s strongest evidence is technical and public: docs, code, benchmarks, changelogs, status pages and community answers. We would expect a tool with deep, accurate public docs and real community use to give AI answers more to work with, though nobody has tested that on DevOps vendors.

What happens to a DevOps vendor that AI answers skip?

It loses trials it never sees, because engineers try whatever the answer names and move on.

  • Silent losses. Product-led funnels start without a form fill. If an assistant names three tools and not yours, the trial never happens and nothing records the loss.
  • Agents that can’t read you. If documentation is hard to parse, an agent may fall back on other sources or other tools. We describe this in what happens when AI agents can’t read your site.
  • A single mention can vanish. Our consistency study repeated each question to ChatGPT five times, and only 25.2% of the brands ChatGPT named were there every time, a warning for any pipeline tool counting on one good answer.
  • Expansion depends on adoption. Fewer teams adopting today means fewer accounts that grow into six- and seven-figure contracts later, we infer.

What does GEO look like when your buyers read docs before marketing?

It makes your docs, code and benchmarks easy for engineers and agents to find, with no guarantee of a recommendation.

  1. Treat documentation as the front door. Keep it public, complete, versioned and fast. Give clear install steps, limits and pricing. Offering clean text versions can help agents; in our llms.txt study, 11.5% of top sites published a valid file, but no study yet shows that AI answers use it. See whether agents recommend readable websites.
  2. Ship working examples. Maintain example repositories, templates and integration guides on GitHub for the stacks buyers use. Data tools face the same stack question; see how a database gets picked for the stack.
  3. Write honest comparison and migration guides. Engineers ask “X or Y” and “how do I move from Z.” Answer with real trade-offs; see whether comparison pages help.
  4. Earn community and third-party proof. Conference talks, foundation projects, independent benchmarks and technical publications carry more weight than your own pages. Take part in communities honestly; planted posts backfire, as we explain in legitimate GEO versus manipulation.
  5. Keep facts consistent. Pricing, limits and compliance status should match across docs, pricing pages and marketplaces; see how to fix wrong brand information in AI answers.
  6. Watch both assistants and agents. Ask your tooling questions repeatedly in ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity, Copilot and Claude, and check your docs logs for agent visits; how many prompts to track sets the sample size. The wider software approach, without the developer-first twist, is in B2B SaaS revenue from AI search.

What is still unknown about AI search and DevOps tool adoption?

No one has measured whether AI mentions lift a DevOps vendor’s pipeline or its net retention.

  • Few DevOps-specific studies. Most AI search research covers software in general, not CI/CD or observability. Engineering software has the same gap, with no public survey of how engineers choose CAD, simulation or PLM tools, as our guide to CAD, simulation and PLM software notes.
  • Vendor data. Mintlify, Vercel and GitLab report their own numbers, and Stack Overflow sells advertising to tool vendors.
  • Agents are new. How coding agents choose tools, and how stable those choices are, is barely studied.
  • Revenue effects remain unproven for tools engineers adopt. Does AI visibility drive business results weighs the general evidence.

How can a DevOps vendor find out whether AI answers are losing it trials?

Ask the questions engineers ask, read what assistants and coding agents return, then fix the docs behind it.

Start from what a platform engineer or SRE would ask at each stage: category, alternatives, comparisons, migration, standards and pricing. Put each question to ChatGPT, Gemini, Claude and the other assistants several times, and point a coding agent at your docs to see what it can extract. Note which tools come back, which sources are cited, and whether your rate limits, pricing tiers and integrations are stated correctly.

We can do this with your team: talk to us about a docs and AI visibility review. It shows where AI answers put you in front of engineers, and which gaps in your documentation and third-party proof are most likely costing you free-tier signups, team adoption and platform contracts. Our generative engine optimization service page describes how the ongoing work runs for engineer-led products, from documentation and example repositories to repeated checks in assistants and coding agents.

Frequently asked questions

Do engineers trust AI tool recommendations?

Not fully. 46% of developers distrust AI accuracy, so they check docs and peers before adopting a tool.

Is documentation more important than marketing pages for AI visibility?

Probably, for DevOps. Documentation is developers’ top learning resource (68%), and agent reads of docs are growing fast on Mintlify-hosted sites.

Does Reddit matter for DevOps tools?

For Google’s AI Overviews, clearly: 35.4% of those on B2B software searches in our study cited a Reddit thread.

How long before AI visibility shows up in revenue?

Trials can start within days; enterprise contracts follow team adoption, which can take quarters.

Should we publish an llms.txt file?

It is cheap, but unproven: 11.5% of top sites publish one, and no study shows AI answers use it.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work