Blog

Social Listening With Trend Data in Claude

Run social listening inside Claude with TrendsMCP: pull live boards from X, TikTok and Reddit, check growth with get_growth, and see what November 17 showed.

Trends MCP team Published Data as of 2026-11-17

Line chart of weekly Google Search interest for social listening from August to November 2026, peaking at 74 in late September and falling to 28 by November 14

Social listening with trend data means asking an AI assistant to watch live boards and search interest, then to test whether a topic is rising or fading before acting on it. Inside Claude with TrendsMCP, that takes four of the five MCP tools: get_top_trends for the live boards, get_growth for the direction of travel, get_time_series for the history behind a spike, and get_context for what people are actually posting. A run on November 17, 2026 showed the split clearly. X and TikTok boards were fresh that morning, Reddit data was 21 days stale, and Google Search interest in "social listening" had fallen 45.1% in three months: the index stood at 51 on August 15 and at 28 on November 14. The exact phrase "social listening with trend data" has too little search volume for Google to report, while the parent term "social listening" draws 14,800 monthly searches.

One board per platform beats one feed for everything

Each platform answers a different listening question, so the run starts with a separate get_top_trends call per board. X shows what is loud right now, TikTok shows what is spreading through short video, and Reddit shows what communities are discussing at length. Mixing them into one list hides the difference: a topic can lead on TikTok while barely registering on X, and that gap is the signal.

The call takes a board type and a limit. Ten rows cost one request per board on MCP, so a three board sweep costs three requests:

{"mode": "get_top_trends", "type": "X (Twitter) Trending", "limit": 10}
{"mode": "get_top_trends", "type": "TikTok Trending Hashtags", "limit": 10}
{"mode": "get_top_trends", "type": "Reddit Hot Posts", "limit": 10}

In Claude the same calls happen by asking for the top 10 on each board in plain words. The prompt that works reads: "Pull the top 10 from the X Trending, TikTok Trending Hashtags and Reddit Hot Posts boards, then tell me which topics appear on more than one board." Asking for the cross board overlap in the same prompt keeps the answer short and forces the comparison instead of three separate dumps. Run on November 17, the overlap was one theme: American football, with five NFL entries in the X top 10 and nfl at number 2 on TikTok. Setup for Claude is covered on the MCP server for Claude page.

The four MCP calls that make up a listening pass

The four tools split the work cleanly: boards find what is moving, growth tests direction, time series gives the shape, and context gives the posts. A listening pass uses them in a recommended order, and each one answers a question the others cannot. get_top_trends finds what is moving now. get_growth tests whether a named topic is rising or fading over 3 and 12 month windows in a single call. get_time_series pulls the weekly history behind the move: 261 weekly points for "social listening", about five years. get_context pulls the posts, videos, stories, comments and markets behind a keyword, one request per source and at most five sources per call.

The growth call for this piece looked like this, with both windows in one request:

{"mode": "get_growth", "keyword": "social listening", "source": "google search", "percent_growth": ["3M", "12M"]}

Cost stays low because each tool charges per call, not per window. get_growth costs one request with every window included, get_time_series costs one request per call, and get_top_trends costs one request per 10 rows. The fifth tool, get_top_posts, needs a Business plan, so it sits outside this pass.

What the boards showed on November 17

The X board on the morning of November 17 was dominated by sport. Eight of the top 10 entries tied to American football or wrestling: Ravens, Chargers, Jacoby Brissett, Justin Herbert and Derrick Henry from the NFL, plus #WWERaw, LA Knight and Becky Lynch from wrestling. The other two were Bond and United Airlines. A listener tracking a sports brand would have found the whole morning there, while a listener tracking anything else would have found almost nothing on X that day.

TikTok told a different story. The top hashtag was tiktokshopblackfriday, followed by nfl, meme, christmas and fortnite, with tiktokshop at number 7 and ad at number 10. Shopping and the holiday season led the board, not sport. The same morning, on the same internet, X was a sports bar and TikTok was a shopping mall. The one theme the two boards shared was American football, and the sharpest difference was shopping: TikTok led with tiktokshopblackfriday while X led with the Ravens.

Reddit was the weak link in this run. The Hot Posts board carried a timestamp of October 27, 2026, 21 days before the check, and its top post was about Angelica Jardine becoming the second woman to lift the Dinnie stones side by side. A three week old board cannot tell a listener what Reddit cares about today. The Reddit source was also unavailable for get_context, so the context call for "social listening" returned zero items with a source unavailable error. Any listening report that quotes Reddit numbers from this setup should flag them as stale, not present them as current.

The three boards summed up: X was fresh at 03:01 UTC with Ravens on top, good for same day sport and event chatter. TikTok was fresh at 09:01 UTC with tiktokshopblackfriday on top, good for shopping and seasonal spread. Reddit was stale, timestamped October 27 with a Dinnie stones lift on top, good for nothing current.

Search interest in social listening fell by almost half

Google Search interest in "social listening" reached 28 on November 14, after standing at 51 on August 15 and at 54 on November 15, 2025. That is a 45.1% drop across three months and a 48.15% drop across twelve months. The related term "social listening tools" moved the same way, only less sharply: it stood at 37 in mid-August and 43 a year earlier, and reached 25 by November 14, drops of 32.43% and 41.86%.

The weekly series behind those numbers has a clear shape. Interest climbed through September, peaked at 74 in the week of September 26, then dropped fast: 47 in the week of October 3, 34 the next week, and 28 by November 14. A listener who only saw the September peak would have read a boom. By November 7, six weeks after the peak, interest was back to 34, below the late August level of 47, so the growth window, not the peak, tells the story.

Keyword Planner adds the demand context. The exact phrase "social listening with trend data" has too little volume for Google to report, which means almost nobody searches it word for word. The traffic sits one level up: "social listening" at 14,800 monthly searches and "social listening tools" at 9,900, both medium competition. A page that targets the exact long phrase still has to earn its readers from the broader terms, so the piece has to answer the broader question, not just repeat the phrase.

What breaks first in this setup

The Reddit gap is the biggest one. A stale board plus an unavailable context source means Reddit listening through this setup is guesswork until the source recovers, and the honest move is to say so in the report and check again the next day rather than quote October numbers as if they were live. The fix is a freshness check in every pass: read the board timestamp first, and skip any board older than a day.

The second gap is coverage. The boards rank public trending topics, so anything behind a login stays outside this pass. A brand name that never trends will not appear on a board at all. The test for that case is to run get_growth on the brand name and on the category term side by side and compare the two series, which shows whether the brand tracks its category or trails it. The REST side of the same data is covered on the best social listening APIs page.

The third gap is the temptation to report a spike as a story. The September peak of 74 looked like a surge, and the seven weeks after it erased the gain, landing at 28 by November 14. The rule that prevents that mistake is simple: never report a board position without the growth window next to it. On November 17, tiktokshopblackfriday sat at number one on TikTok while search interest in social listening itself sat at 28, down from 54 a year earlier. Rank and direction answer different questions.

The single next step is to run the same four calls on a brand name and compare the result with the category term, then check the wider field in the best AI agents for social listening roundup and the plan details on pricing.

Questions

Which MCP tool shows what people are actually saying?

get_context. It returns posts, videos, stories and comments about a keyword from each source, with engagement. One request per source, at most 5 sources per call.

Why did the Reddit board look out of date in this run?

The Reddit Hot Posts board carried a timestamp of October 27, 2026, 21 days before the check on November 17, and the Reddit source was also unavailable for get_context. Treat Reddit numbers from this setup as stale until a fresh board arrives.

Sources