Find Third-Party Mentions of a Brand on the Web and X

Brand Mentions Finder builds a monthly list of third-party pages and public X posts about a brand. It searches the web and X for the last 30 days, then drops the brand's own sites and the brand's own account. Authors of X posts are shown as a user. The rows are what Desearch returned for that search, not a full count of the conversation.

How this differs

The social listening report pulls the last 7 days of X posts and web pages and keeps the brand's own pages in the sheet. The Twitter complaints monitor keeps posts that pair the brand with complaint words and sorts them into themes. This script looks back 30 days, asks for links, and removes the brand's own domains and account before it labels tone.

The problem

A founder or PR person who wants "what others said this month" still has to skip the brand's own site, skip the brand's own posts, and paste the rest into a sheet. Most of a naive search is the brand talking about itself.

What the app does

• Reads the brand, what it is, brand terms, the brand's own account, own domains and a 30-day window from config.json.
• Calls AI search for web and X links, then calls X search for more posts with a likes floor.
• Drops links on the brand's own domains and subdomains, posts from the brand's own account, duplicate links, rows with no brand term, and rows that match another meaning of the name.
• Writes a table with where, link, a short snippet, tone (positive, neutral, negative) and date, plus a "left out" list.
• Shows each X author as "a user".

How it works

1. Links. POST /desearch/ai/search with tools: ["web", "twitter"], result_type: "ONLY_LINKS", streaming: false, and start_date / end_date covering the last 30 days.
2. More posts. GET /twitter with the brand terms inside parentheses, -from: for the account you name, sort=Latest, min_likes and count. Dates are YYYY-MM-DD.
3. Clean. Domain and account filters run after the response. The request does not ask the API to drop domains.
4. Tone. A small word list marks positive, negative or neutral. Read each row before you act on it.

The code

The X query keeps the OR group in parentheses and leaves out the brand account:

def x_query(cfg):
    group = term_query(cfg["brand_terms"])
    account = (cfg.get("own_account") or "").strip().lstrip("@")
    if account:
        group += " -from:%s" % account
    return group

The AI search body:

def ai_payload(cfg, start, end):
    brand = cfg["brand"]
    what = (cfg.get("what_it_is") or "").strip()
    prompt = "Third-party reviews, articles and posts that mention %s" % brand
    if what:
        prompt += ", %s" % what
    prompt += ", from the past month"
    return {
        "prompt": prompt,
        "tools": ["web", "twitter"],
        "result_type": "ONLY_LINKS",
        "streaming": False,
        "start_date": "%sT00:00:00Z" % start.isoformat(),
        "end_date": "%sT23:59:59Z" % end.isoformat(),
    }

Own-domain check:

def on_own_domain(url, domains):
    host = host_of(url)
    if not host:
        return False
    for domain in domains or []:
        d = domain.lower().strip().removeprefix("www.")
        if d and (host == d or host.endswith("." + d)):
            return True
    return False

The two live calls:

def run_live(cfg, session=None):
    key = api_key()
    if session is None:
        import requests
        session = requests.Session()
    session.headers["Authorization"] = key
    start, end = window(cfg)
    payload = ai_payload(cfg, start, end)
    params = x_params(cfg, start, end)
    ai = session.post(BASE + "/desearch/ai/search", json=payload, timeout=120)
    xres = session.get(BASE + "/twitter", params=params, timeout=90)
    raw = {
        "run_at": dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"),
        "cfg": cfg,
        "start_date": start.isoformat(),
        "end_date": end.isoformat(),
        "ai": {"status": ai.status_code, "body": _json(ai), "payload": payload},
        "x": {"status": xres.status_code, "body": _json(xres), "params": params},
    }
    return raw, key

The key is sent only in the Authorization header and is not written to the output.

Endpoints

• POST /desearch/ai/search
• GET /twitter

Example output

The 2026-10-15 sample used a Mac launcher app as the brand. AI search returned web pages and X posts for the last 30 days, and the X search returned 20 posts. 6 of 10 web pages were on the brand's own site and were left out. 1 post in 30 was about something else and was left out. X authors in the public sample are shown as a user.

Run it yourself

The public examples folder linked above includes the script, config.json, a saved prompt and a README. Install its requirements, set DESEARCH_API_KEY from console.desearch.ai, and use the run command in that README. You can also paste the saved prompt into an MCP client pointed at https://mcp.desearch.ai/mcp.

What it costs

Each run makes two calls. The sample write-up does not add the X-Desearch-Cost-Usd headers into one total. See pricing.

Ideas to extend it

• Run it on the first of the month and keep each CSV.
• Add a column for the brand's own pages so you can see what was dropped.
• Split tone counts by week.

{
  "mcpServers": {
    "desearch": {
      "url": "https://mcp.desearch.ai/mcp",
      "headers": {
        "Authorization": "Bearer <DESEARCH_API_KEY>"
      }
    }
  }
}

Use Desearch to find public web pages and X posts from the last 30 days that mention the brand I name. Drop the brand's own domains and the brand's own account. Label tone only from text you retrieved. Show X authors as a user. Do not invent pages or posts.