The finding in one paragraph
We picked 10 funded B2B developer-tools and infrastructure companies, all with strong category positions in their sub-niche. We asked four leading LLMs (ChatGPT, Claude, Gemini, Perplexity) five buyer-research questions any engineering leader or platform team lead would plausibly type into an AI assistant. We ran each question 3 times per engine for variance. That produced 600 distinct LLM responses. Only 13 of those responses mentioned any of the 10 target brands by domain or by name: 8 mentioned LlamaIndex, 3 mentioned MacStadium, 2 mentioned Qase, and the remaining seven brands appeared zero times. The category is being defined by a different set of brands entirely.
Where each brand landed
Seven of ten scored 0 / 60. Three broke through. The leader (LlamaIndex) earned 8 mentions out of 60 LLM responses. Even the top brand was cited less than 14% of the time.
| Brand | Domain | Category | Score |
|---|---|---|---|
| LlamaIndex | llamaindex.ai | LLM framework and RAG infrastructure | 8 / 60 |
| MacStadium | macstadium.com | Mac cloud + CI infrastructure | 3 / 60 |
| Qase | qase.io | QA test management | 2 / 60 |
| ActiveState | activestate.com | Open-source supply chain security | 0 / 60 |
| HumanSignal | humansignal.com | Data labeling for ML teams | 0 / 60 |
| FusionAuth | fusionauth.io | Authentication and identity for developers | 0 / 60 |
| Liquibase | liquibase.com | Database change management and CI/CD | 0 / 60 |
| Network to Code | networktocode.com | Network automation | 0 / 60 |
| RapidCanvas | rapidcanvas.ai | AI/ML platform and applications | 0 / 60 |
| Synergis | synergis.com | Engineering document management | 0 / 60 |
Who LLMs cite instead
The same 600 responses cite the brands below by name, repeatedly. These are the companies that have entered the AI-search conversation for devtools buyer prompts.
- LangChain82
- TestRail75
- Auth055
- Okta55
- GitHub Actions51
- CircleCI49
- Xcode Cloud49
- Jenkins40
- Jira37
- Zephyr36
- PropelAuth36
- Haystack36
- Bitrise35
- PractiTest35
- Fastlane31
- Zephyr Scale31
Mention counts are bolded-product-name references across all 600 LLM responses.
Methodology
This study was designed to be replicable. The exact prompts, models, settings, and matching logic are documented below. Anyone with API access to the four engines can re-run it.
Anthropic Claude (claude-haiku-4-5-20251001)
Google Gemini (gemini-2.5-flash-lite)
Perplexity (sonar)
max_tokens 800
N = 3 trials per (engine, prompt)
The 5 standardized buyer prompts
- What is the best macOS CI/CD infrastructure for iOS and macOS app development teams?
- How do enterprises secure their open-source software supply chain in 2026?
- What is the best authentication and identity platform for B2B SaaS developers?
- What tools do QA teams use for test management and release coverage at scale?
- What is the best LLM framework for building production RAG applications?
The same 5 prompts were applied to every brand audit. This is the only way to make a category comparison defensible: if each brand got its own custom prompts, the rankings would reflect the prompt selection, not the brand visibility.
What this means if your devtools brand is on the invisible list
Most funded B2B devtools vendors have not entered the AI-search conversation. They built websites optimized for SEO and developer-doc indexing, not for how LLMs synthesize answers. When a platform engineer or CTO asks an AI assistant a category question, the answer comes back with the brands that show up in Stack Overflow answers, in GitHub READMEs that LLMs ingested, in YouTube engineering content, and in long-form analyst writeups. The dominant set is not necessarily the best tool. It is the set the LLMs have read about most often.
The fix is mechanical. LLM-friendly content follows specific patterns: structured FAQ content matching exact buyer-prompt phrasing, schema-marked product pages, citation footprints on third-party developer sites the LLMs already index, and content that answers comparison queries directly. LlamaIndex did some combination of these things, intentionally or not, and earned a 13% citation share where the median brand earned 0%.
Want to know if your devtools brand is invisible?
Web Cited runs the same measurement against your domain, your buyer prompts, and the engines your developer audience actually uses. The Audit comes with a click-to-copy Playbook your engineers ship from in the next sprint. Five business days. Fixed price.
Order an audit