Ask the assistants your buyers already ask — ChatGPT, Claude, Gemini, Perplexity, DeepSeek — the questions they type when they are choosing. SeenByLLMs counts where your brand shows up, ranks you against the names you compete with, and hands you a report you can send to a client.
Run a check No account, no server. Your OpenRouter key stays in your browser.
Here are the options I would look at first.
Buyers increasingly ask ChatGPT and other assistants for recommendations. If the model names your competitor and not you, that lead never shows up in Google Analytics — you cannot see the loss, and you cannot fix what you cannot measure.
Same scoring as the app. Free OpenRouter model + web search. Eleven buyer questions each. Numbers below are from these runs — not demos, not invented uplift.
4 / 11 questions with Plausible absent · SOV 33% vs Matomo 19%
Example loss: “Which privacy analytics tools work well with Next.js?” → models named Simple Analytics, not Plausible.
3 pitch targets after excluding competitor/own/platform domains (beam-privacy.com, docs.oursprivacy.com, privatusanalytics.com — docs.simpleanalytics.com and github.com stay in the citation list). 35 distinct cited domains. Run cost $0.07 in 0.7 min (11 answers — 11 tabs if done by hand).
3 / 11 questions absent · SOV 18% vs ConvertKit 23% (ConvertKit 10 mentions vs Buttondown 8)
Example: “Which email newsletter tools do you recommend for a solo creator?” → ConvertKit named, Buttondown not.
10 pitch targets (best4beginners.com, creatormailstack.com, … — kit.com / beehiiv.com excluded as competitor domains). 41 domains. $0.08 · 0.8 min.
3 / 11 questions absent · SOV 21% vs Typeform 24%
Example: “Compare no-code form builders for startups” → Typeform named, Tally not. Same on product-feedback surveys.
9 pitch targets (formester.com, fabform.io, formbricks.com, … — paperform.co excluded as competitor). 44 domains. $0.08 · 0.9 min.
Also measured: Cal.com (absent 2 / 11, score 70, 3 pitch targets, $0.08) and Raycast (absent 2 / 11, score 67, 2 pitch targets, $0.08). Read all five case studies: method, every answer, cited pages, and what each brand could fix.
Create a key at openrouter.ai/keys, set a spend limit, paste it in the app (or OAuth). The key stays in your browser — we never host it.
Your product name, up to five rivals, and the questions buyers actually ask assistants. Or start from suggested prompts and edit.
Which questions omit you, who got named instead, which URLs cite only competitors, and prioritized actions — as HTML or CSV.
These five showcase runs (11 questions × 1 free model + web search each) cost about $0.38 total on OpenRouter — roughly $0.07–$0.08 per brand, under a minute each. That is 11 model answers per brand (11 questions × 1 model) that you would otherwise open and score by hand.
Competitor AI-visibility tools from our 2026-10-16 research start around $29/mo (Otterly Lite, Elmo Cloud) and $79/mo (LLMrefs), with Peec from $95/mo. SeenByLLMs itself is $12/mo or $49 lifetime early-bird; model calls are billed to your own OpenRouter key. We do not claim traffic or revenue uplift we have not measured.
Monthly $12, or lifetime $49 for the first 100 buyers (then $79). You still pay OpenRouter for the model calls. A 20-question run on the five default models is about $0.30.
Buy monthly Buy lifetime Already bought? Activate
$12/month
$49
Early-bird for the first 100. Then $79 once.
In this browser. SeenByLLMs has no server and no account. Export the report before you switch machines, because nothing syncs.
The key is stored in this browser's local storage and is sent to OpenRouter and nowhere else. Nothing is proxied through us, because there is no us to proxy through. Give the key a spending limit so a mistake cannot cost more than you meant.
Score = round(100 × (0.5 × mention rate + 0.3 × share of voice + 0.2 × rank score)), where rank score is 100 for first place down to 0 at sixth or worse. Failed calls are left out of every number. The formula is published because the other tools in this market keep theirs secret.
Plain text matching on your brand name, the aliases you listed, and your competitors. It costs nothing extra and the same answer always scores the same way. Ranking reflects the order companies first appear in an answer.
Web search is on by default: each question is sent with OpenRouter's openrouter:web_search
tool, so answers come back with the pages behind them, and every source is labelled by what kind of
site it is — your own, a competitor, a review site, a forum, or editorial. It costs about $0.007 per
request on models that do not search natively, billed by OpenRouter to your account. Turn it off and
the answers come from the model's memory, with no sources to show.
They are different models with different training data and, for some, different search results. That disagreement is the finding. A brand that only Gemini names has a different problem from one that nobody names.