BrandMentionWatch docs
Set up in 5 minutes: create a brand, add keywords and competitors, connect sources, set alert channels.
1. Sources
We use only official APIs, open datasets and publisher RSS — never prohibited scraping.
| Source | Status | Notes |
|---|---|---|
| Coming soon | Reddit Data API via OAuth. Commercial use needs a Reddit Data API agreement (founder step). | |
| RSS / Atom (news, blogs, forums) | Live | Publisher RSS feeds are offered for syndication; we store title/summary + link only. |
| News (GDELT, 65 languages) | Live | GDELT is an open dataset; free for any use with attribution. |
| Hacker News (forum) | Live | Public HN Search API by Algolia. |
| YouTube (videos + comments) | Coming soon | YouTube Data API v3 (10k units/day free). search.list = 100 units; commentThreads = 1 unit. |
| X (Twitter) | Coming soon | X API v2 recent search (7 days). Paid tier required for volume (founder step). |
| Apple App Store reviews | Live | Apple's public customer-reviews RSS feed. |
| Google Play reviews (own app) | Coming soon | Google Play Developer API reviews.list (last 7 days, own apps only). |
| Google reviews (Business Profile) | Coming soon | Google Business Profile API (owner-authorised). Supports replying. |
| Podcasts (Apple directory + show feeds, transcripts) | Live | Apple's public iTunes Search API + publishers' own RSS/transcript files. Audio is transcribed on our GPU, not stored. |
| Facebook + Instagram (Meta Graph API) | Coming soon | Official Graph API, owner-authorised. Hashtag search limited by Meta to 30 hashtags per 7 days. |
| Support email | Forwarding address from us | POST /api/inbound/email → unified inbox |
2. Noise guard
Every mention gets a relevance score. Whole-word keyword hits score 1, partial hits 0.4, and anything containing an exclude keyword is dropped before it reaches you. Example: track “Brewly”, exclude “brewly beer”.
3. Sentiment & languages
A built-in lexicon scores sentiment with negation handling. Where the lexicon is unsure, our own Qwen model on Surge GPUs refines it; Claude is used only as a fallback. You can correct any sentiment in the inbox.
4. Alerts
- Spike — last hour’s volume is ≥ N standard deviations above the 48-hour normal (min 5 mentions).
- Negative surge — negative share in the last 6 hours is ≥ N points above the 7-day baseline.
- Keyword — e.g. “lawsuit”, “fraud”, “recall”.
- Influencer — a new mention from an author with influence score ≥ N (0–100: reach + engagement + source authority).
Channels: in-app always; Slack and Teams via incoming webhooks; email via Resend.
5. Inbox & SLAs
Priority is set automatically: urgent (very negative from a big account, or trust issues like “fraud”) 1h · high 4h · normal 24h · low 72h. Today a reply is logged in the inbox and you post it on the platform via the link (email tickets are answered by email). Posting replies straight to Google reviews, YouTube, Reddit, Facebook and Instagram is coming soon.
6. REST API
curl -H "Authorization: Bearer sl_xxx" \ "https://brandmentionwatch.com/api/v1/mentions?brandId=1&since=2026-10-01&sentiment=negative"
Returns mentions with sentiment, language, topics, influence and competitor tag. Create keys in Settings → API keys.
7. Plan limits
Mentions count per workspace per calendar month (IST). At 80% and 100% you get an alert. At the cap, new social and news mentions pause until the 1st — reviews and support tickets always arrive. Need more? Add +10,000 mentions for ₹999/month (Brand24 charges $25), or upgrade. Brands, keywords (competitor terms count) and users are checked when you add them.
8. Podcasts
We search the Apple Podcasts directory for episodes that name you, and read the feeds you add. When a show publishes a transcript (Podcasting 2.0 podcast:transcript), we find where you are spoken about and show the timestamp, e.g. [12:01] “I switched to Brewly last month…”. Transcribing shows that publish no transcript is coming soon.
9. Ask your mentions
Ask in plain words — “why are payments failing?”, “refund kab milega” — and get an answer that cites the mentions it used. Retrieval is keyword-based over your own data; our Qwen model writes the answer, and an answer citing a mention that wasn't retrieved is rejected.
10. Review campaigns
Each customer gets a personal link. Everyone sees the same public review button (we never filter unhappy customers away — that breaks Google’s review policy); private feedback is optional alongside.