Every developer reading this knows what Sentry does. It is the default answer to "how do I track errors in production", and has been for years.
We asked seven AI answer engines "what is the best error monitoring tool for a small team". They cited scoutapm.com, observeone.com, guideflow.com and buildmvpfast.com.
Sentry was not mentioned. Not by one engine, on any run.
The list
We ask 100 buyer questions across 10 categories to Google AI Overviews, Google AI Mode, ChatGPT, Claude, Gemini, Perplexity and Copilot, and publish the results at seobuilder.tech/ai-citations. As of today that is 2,586 citations across 1,435 distinct domains.
Searching those 1,435 domains for names any developer would recognise:
| Company | Times cited |
|---|---|
| Sentry | 0 |
| Supabase | 0 |
| Railway | 0 |
| Neon | 0 |
| PlanetScale | 0 |
| Resend | 0 |
| Linear | 0 |
| Tailscale | 0 |
| Cal.com | 0 |
| Clerk | 2 |
| Vercel | 8 |
| Plausible | 6 |
| Render | 1 |
| Fly.io | 1 |
| PostHog | 1 |
Substring match, so a citation of docs.supabase.com or blog.sentry.io would
have counted. None existed.
We checked that we asked the right questions
A domain missing from an answer means nothing if you never asked a question where it belonged. So here are the ten developer questions we ask, with who actually got cited.
| Question | Cited |
|---|---|
| best database for a new saas app | reddit, stackoverflow, pingcap.com, buildmvpfast.com, whipp.studio |
| best authentication service for a small saas | reddit, auth0.com, kinde.com, propelauth.com, descope.com, supertokens.com |
| best error monitoring tool for a small team | reddit, scoutapm.com, observeone.com, guideflow.com, buildmvpfast.com |
| best way to host a node.js api | reddit, youtube, fly.io, render.com, heroku.com |
| which vector database should i use for rag | reddit, zenml.io, braintrust.dev, digitalocean.com, firecrawl.dev |
| best ai coding assistant for a solo developer | reddit, vellum.ai, augmentcode.com, zapier.com, collabnix.com |
| should i use next.js or astro for a marketing site | reddit, cosmicjs.com, contentful.com, vercel.com |
| is tailwind css worth using on a new project | reddit, quora, glideapps.com, devot.team |
| react native or flutter for a first mobile app | medium, reddit, sendbird.com, quora |
| is aws cheaper than a paas for a small app | pandastack.io, dotsquares.com, massivegrid.com, vps.do |
The database question is asked. Supabase, Neon and PlanetScale are not there. The auth question is asked, and it names five auth vendors. Clerk is not one of them. The hosting question is asked and names three PaaS providers. Railway is not one of them.
Two things this actually shows
Reddit is cited for 71 of the 99 questions. It is the single most cited domain in our corpus by a distance, and on developer questions it is close to universal. If you want to be in these answers, the thread where someone asked your question three years ago matters more than your documentation.
The named winners are sites you have never heard of. scoutapm.com, observeone.com, guideflow.com, buildmvpfast.com, whipp.studio, devot.team. These are not household names in developer tooling. They are, in most cases, agency blogs and content sites that wrote a listicle.
Brand recognition among developers and citation by an answer engine turn out to be close to unrelated. A company can be the obvious answer to a question inside the industry and be absent from the machine's version of that answer.
What we are not saying
We cannot tell you why any single domain was absent. We tested the usual explanations on our own data and they did not hold: schema markup, list counts and heading counts did not separate cited pages from uncited ones once we controlled for page length. Anyone selling you a checklist for this has not tested it.
We also are not saying these companies did anything wrong. Sentry does not have an SEO problem. It has, on this evidence, an answer engine problem, and those are different things with different fixes.
Some honest limits:
- 99 questions in 10 categories, one run. A different question set produces a different table. Ours is public, so you can see exactly what we asked.
- Engines are not deterministic. Ask the same question twice and the cited sources can differ. We recorded one answer per engine per question.
- US English, one location. Results elsewhere will differ.
- A citation is not a visit. We measured whether an engine named a domain. We did not measure whether anyone clicked, and neither can the engines.
If you work at one of these companies
The board is public and updates on a rotation, so you can check your own position without talking to us: seobuilder.tech/ai-citations.
If your domain is in the 1,435 we recorded, you are already ahead of most. 1,244 of those 1,435 domains were cited for exactly one question, which tells you how long the tail is and how little it takes to be on it.