FAQ
Frequently asked questions
Clear answers about wallet credit, usage, subscriptions, and how Tycoon charges for work.
Most brand monitoring tools have terrible signal-to-noise. How is this different?
Generic brand monitoring surfaces every mention and leaves prioritization to you — which means you spend 20 minutes a morning scrolling and acting on maybe 2 items. Tycoon's AI CMO does the prioritization up front: a mention from an account with 180 followers in a subreddit that's basically dead gets logged but not escalated; a mention from a founder with 50K followers in a relevant pro community gets flagged with a draft reply within 10 minutes. The ratio of 'things I look at' to 'things that matter' goes from 1:20 to roughly 1:1. Noise is still captured for search history — you can look up any mention later — but it doesn't interrupt your day.
How do you handle brands with ambiguous names — like if my product is called 'Tycoon' and there's also a game called 'Tycoon'?
Name disambiguation is the first classification step. The AI uses context clues: words in the mention (SaaS / AI / game / MMO), the surrounding thread topic, the poster's bio and history, and your product's taxonomy. For highly ambiguous names it maintains an exclusion filter (ignore mentions of your name when paired with 'MMO' or 'Xbox'), tunes it continuously based on your overrides, and errs toward false positives rather than false negatives (it's cheaper to ignore a false hit than miss a real one). Over ~2 weeks of training on your corrections, precision converges to ~95% on normal brand names and ~85% for especially ambiguous ones. For very generic brand names (e.g. 'Launch', 'Pitch') you may need tighter filters specified up front.
Can it actually monitor LLM citations? ChatGPT answers change every time you ask.
LLM citation is sampled, not exhaustive. The AI Head of Growth runs a fixed prompt panel — say, 50 prompts relevant to your product — across each major LLM (ChatGPT, Perplexity, Claude, Gemini) every week. Each prompt runs 3-5 times to average out variance. Share-of-voice is computed as (# of prompts where you're cited) / (# of prompts total). This gives you a trend line: week 1 you were cited in 14/50 prompts, week 8 you're cited in 27/50. Competitor SOV runs on the same prompts. It's not real-time alerting on LLM mentions (that's not possible with any tool), but it is a reliable longitudinal signal of your AI visibility, which is increasingly the main distribution channel for B2B SaaS.
What about private communities — Slack, Discord, private podcasts, etc? I'm more worried about a Slack community roast than a Twitter one.
Public monitoring can't reach those, but you can bridge them in two ways. One: join the communities as yourself (or have your AI Customer Support join as a clearly-labeled AI team member in communities that allow it), and Tycoon ingests the member's relevant messages. Two: customers who want to report something from a private community can forward the message to a dedicated inbox (brand-alerts@yourco.com) and the AI triages it the same way. Neither gives you full coverage of private spaces, but in practice the private-community mentions that matter tend to get surfaced to you eventually through one of these paths. The AI is honest that private-community coverage is incomplete and flags it in the weekly report.
How does this avoid auto-replying to a tweet and making the AI-voice problem worse?
Auto-send to public platforms is off by default and not configurable on Twitter/Reddit/HN. Every public reply requires your one-tap approval. The AI drafts, labels the reply type ('response to complaint', 'thank-you to advocate', 'clarification to misinformation'), attaches the original context, and waits. What auto-sends: private channels where you've opted in (Intercom ack to a bug report, DM response to a product question from someone already in conversation with support). Public posts always go through you. This costs some latency on weekends when you're slow to approve, but it keeps the brand voice unmistakably human-run. A single bad public AI reply damages trust more than a dozen good ones help it.