What 126,814 citations tell us about AI search for dev tools
Five findings from DevTune’s State of AI Search for Dev Tools 2026
Today, we published the first State of AI Search for Dev Tools report.
We analysed 126,814 citation records across 1,075 representative tool-evaluation scenarios, 43 developer-tool categories, and five AI search platforms. Each question was tested weekly during the 90-day reporting period.
We wanted to understand which pages AI search actually uses, why certain brands appear more consistently than others, and what dev-tool teams should do in response.
Here are five of the most important findings.
1. Detailed content gets cited more than conventional product pages
Blog posts were the largest citation surface, accounting for 34.3% of citation records. Documentation was second at 17.3%.
By comparison, product pages accounted for 7.5% and landing pages 2.6%.
This doesn’t mean the homepage is unimportant. The homepage establishes what a company is, who it serves, and how it positions itself. But answer engines often need more detailed evidence before they can use or repeat a claim.
That evidence tends to live in category explainers, implementation guides, documentation, comparisons, integration pages and technical tutorials.
The practical implication is that every important product claim needs somewhere deeper to be proven.
2. There is no single AI-search playbook
The five platforms showed materially different sourcing behaviour.
On Google AI Mode, 64.4% of citation records appeared within the first three source positions. On xAI Search, that figure was 12.7%.
That means Google AI Mode tends to concentrate its evidence among a smaller set of primary sources. Appearing as one of those sources requires a page strong enough to carry a meaningful part of the answer.
Other platforms draw from broader source sets. There, visibility depends more heavily on having consistent evidence across your own site, editorial coverage, community discussions, video and other third-party pages.
A blended “AI visibility score” can conceal these differences. Teams need to monitor each platform separately and understand the particular source gap they are trying to close.
3. Community sources remain an important part of dev-tool discovery
Community sources represented 10.5% of citation records. They were not the majority, but they were too significant to treat as peripheral.
Reddit was the most-cited third-party domain in the entire dataset, receiving 10,831 citations across 5,109 URLs. GitHub and YouTube were also recurring sources.
This reflects how developers already evaluate tools. They look for first-hand experience, implementation detail, limitations and evidence that does not come exclusively from the vendor.
The answer is not to flood communities with promotional posts. It is to participate genuinely, maintain a useful GitHub presence, support accurate independent content and understand how developers describe your product when you are not controlling the page.
4. Being cited is not the same as being recommended
Comparison pages and “best tools” lists collectively accounted for 14.5% of citation records.
These formats clearly influence AI answers, but publishing one does not guarantee that the answer engine will recommend the publisher.
Across 668 vendor-authored ranked pages, the publisher appeared in the resulting answer 74.4% of the time. In 17.1% of page-and-answer pairings, another tracked brand appeared while the publisher did not. For “best tools” lists specifically, that rose to 22.1%.
A vendor can therefore publish a page that earns citations while giving an answer engine the evidence and competitor names it needs to recommend someone else.
Useful comparison content should still be honest. It should define clear criteria, explain trade-offs and say who each product is suitable for. But teams need to measure citations, brand mentions and recommendations as separate outcomes.
5. Category leadership can become highly concentrated
AI-search visibility was not evenly distributed within categories.
Northflank appeared in 46.4% of AI Code Sandboxes & Agent Runtimes answers, compared with a category median of 5.7%. That placed it 40.8 percentage points above the typical tracked brand and 22.7 points ahead of second-placed Modal.
Other categories were much more contested, while some had no strongly visible leader at all.
This is why broad benchmarks are useful for understanding patterns, but individual decisions require category and brand-specific monitoring. The opportunity facing an established category leader is fundamentally different from the opportunity facing a less-authoritative brand in an unsettled market.
Read the report
The full report includes platform comparisons, content-type analysis, category benchmarks, brand scorecards, third-party source data and an interactive strategy section.
It also features perspectives from founders and leaders across growth, content, SEO, documentation and developer relations. I’m grateful to everyone who challenged the analysis and contributed their experience.
Explore the State of AI Search for Dev Tools 2026:
https://stateofaisearch.dev/2026
Book a free 1:1 consultation call with me - I keep a handful of slots open each week for founders and product growth leaders to explore working together and get some free advice along the way. Book a call.
View your free public Dev Tool AI Market Presence Report - 500+ dev tools across 47+ verticals and growing, all in the Dev Tool AI Search Landscape.
Sponsor this newsletter - Reach over 11,000 founders, leaders and operators working in product and growth at some of the world’s best tech companies including Paypal, Adobe, Canva, Miro, Amplitude, Google, Meta, Tailscale, Twilio and Salesforce.
