Creator Search Insights API: what exists in 2026

TikTok has no public Creator Search Insights API. The tool lives in the app and at tiktok.com/csi, and nothing in TikTok's developer documentation exposes its search numbers. The way to get them from code is TokConnect: three of its 34 research tools return the same topic records, with 7-day search counts and video counts for any phrase, and a first result takes minutes.

Neither the Research API nor the Business API returns search topics, and no partner tier adds one. Lower down, for developers: the runnable Python, the JSON it returns, a field reference for every value in a topic record, how to tell a phrase people type from one of TikTok's generated topic groups, and the follow-up calls that expand one topic into a cluster.

Share or embed this infographic

Free to use on your site, in a deck or a newsletter with a credit link to TokConnect. Copy the code below.

<a href="https://tokconnect.com/guides/creator-search-insights-api/"><img src="https://tokconnect.com/graphics/creator-search-insights-api/summary.png" alt="Is there a Creator Search Insights API? Not from TikTok: 0 TikTok endpoints we found for search demand; 3 of 34 TokConnect tools return CSI search data; 10 of 10 sourdough starter records were topic groups; 456K 7-day searches for Sourdough cookies, a typed search" width="1200" height="630" style="max-width:100%;height:auto"></a><br>Source: <a href="https://tokconnect.com/guides/creator-search-insights-api/">Creator Search Insights API: Does TikTok Have One? (2026)</a> on TokConnect

Copy embed codeDownload PNG

Why there is no Creator Search Insights API

Creator Search Insights is a creator-facing product, not a developer one. TikTok's developer portal is organised around four surfaces, and none of them touches search demand.

The Research API comes closest and still cannot answer the question. Its video query specification filters by keyword, hashtag, region and date, and returns videos. It has no concept of a search topic, a 7-day search count or a content gap. You can count how many videos mention sourdough. You cannot learn that "Sourdough cookies" was searched 455,736 times in the 7 days to 5 October 2026.

One 2026 developer blog, Phyllo's, says a "Creator Search Insights API" is now available. It links no TikTok documentation, and the data it describes (followers, engagement rate, audience demographics per creator) is creator profiles, not search demand. We found nothing matching it on TikTok's developer site.

The broader map of every TikTok API and what each one is good for is in the TikTok API guide.

For developers: the TokConnect tools that return topics

TokConnect exposes the Creator Search Insights topic records, plus ordinary TikTok search, over the Model Context Protocol. It is one HTTP endpoint speaking Streamable HTTP and JSON-RPC.

Get your API key API docs

TokConnect has two doors. Over MCP you open a session and call demand_find_topics. Over REST (any paid plan) you call one URL:

curl "https://api.tokconnect.com/v1/searches?keyword=sourdough%20starter&limit=10" -H "X-API-Key: $TOKCONNECT_API_KEY"

Both cost 1 credit per call; the REST route is documented in the demand API docs. Over MCP, any MCP client library handles that sequence for you, which is the point of using one. If you want the same thing in a chat window rather than a script, the connect pages cover Claude, Codex, Cursor and Gemini CLI, and the TikTok MCP guide explains the protocol.

A Python example that runs

This uses the official mcp package. Install the 1.x line this example was run on with pip install "mcp<2", put your key in TOKCONNECT_KEY, and run it.

import asyncio
import json
import os

from mcp import ClientSession
from mcp.client.streamable_http import streamablehttp_client

URL = "https://mcp.tokconnect.com/mcp"
HEADERS = {"Authorization": f"Bearer {os.environ['TOKCONNECT_KEY']}"}


async def main():
    async with streamablehttp_client(URL, headers=HEADERS) as (read, write, _):
        async with ClientSession(read, write) as session:
            await session.initialize()

            tools = await session.list_tools()
            print([t.name for t in tools.tools])

            result = await session.call_tool(
                "search_topics",
                {"keyword": "sourdough starter", "limit": 10, "language": "en"},
            )
            data = json.loads(result.content[0].text)

            for topic in data["topics"]:
                print(
                    f"{topic['searchVolume']:>10,}  "
                    f"{topic['videoCount']:>6}  "
                    f"{topic['queryText']}"
                )
            print("next offset:", data["nextOffset"])


asyncio.run(main())

import httpx2
from mcp.client.streamable_http import streamable_http_client

async with streamable_http_client(URL, http_client=httpx2.AsyncClient(headers={"Authorization": f"Bearer {KEY}"})) as (read, write):
    async with ClientSession(read, write) as session:
        ...

Two things to note. Tool results arrive as text content, so you parse result.content[0].text as JSON rather than reading a typed object. And list_tools() is worth calling once in development: the tool set grows, and the schemas it returns are the authoritative argument list.

The same search_topics call, made on 5 October 2026, printed this:

 2,923,707       0  dry sourdough starter and dough
 3,755,175     544  sourdough starter making and uses
 4,418,725       0  sourdough eating and starter care
 4,685,193       0  sourdough baking and starter care
 4,027,202       0  Vic sourdough starter and recipes
 3,216,221       0  sourdough starter and baking guides
 4,024,606     273  sourdough starter recipe guides
 4,100,009       0  sourdough baking recipes and starter
 4,220,933       0  sourdough recipes and starter guides
 3,146,555     538  day by day sourdough starter guide
next offset: 10

Data: TikTok Creator Search Insights via TokConnect, 5 October 2026 (the 19 September run returned the same kind of list). Open the snapshot.

Every one of those ten is a TikTok topic group: a generated phrase whose pooled figure covers many related searches, not a search for that wording. Their daily series give them away. Each started the week under 5% of where it ended, the same climb from a few thousand to a few million that every group shows. If your code treats these as keywords, it will report demand nobody typed.

Separating typed searches from topic groups

Add a filter before you use the numbers. This is the rule we use on every page, applied to the records the call returns:

def is_topic_group(topic):
    volume = topic.get("searchVolume") or 0
    trend = topic.get("trend7d") or []
    if volume >= 5_000_000:
        return True
    if volume >= 1_000_000 and trend and trend[-1] and trend[0] < 0.05 * trend[-1]:
        return True
    return topic.get("categoryL1") == "TikTok's Featured Content"

typed = [t for t in data["topics"] if not is_topic_group(t)]

Run over a broader keyword, "sourdough" with limit 20, the filter kept 7 typed searches and dropped 13 groups. These are the ones worth putting in a caption or a dashboard:

Dividing searchVolume by videoCount is a quick way to rank them for content. "extra sour sourdough" had about 1,736 searches per matched video; "will it sourdough" had 74.

The JSON a topic search returns

Here is one full record from the 19 September call plus the envelope around it, trimmed to a single topic:

{
  "topics": [
    {
      "queryId": "7668817079025287188",
      "queryText": "sourdough baking and starter care",
      "popularity": 87,
      "searchVolume": 4685193,
      "videoCount": 0,
      "trend7d": [9597, 106468, 291212, 442767, 1091572, 1611616, 4685193],
      "trendScore7d": [45, 70, 74, 78, 85, 85, 87],
      "categoryL1": "Gourmet",
      "categoryL2": "Food Tutorials",
      "isFavorited": false,
      "channel": "search",
      "tab": "SEARCH"
    }
  ],
  "hasMore": true,
  "nextOffset": 5,
  "count": 5,
  "channel": "search",
  "tab": "SEARCH"
}

That record is a topic group: trend7d starts at 9,597 and ends at 4,685,193, so its first reading is under 5% of its last. Its searchVolume pools many related searches, and videoCount: 0 on a group is not a content gap you can fill with one phrase. A typed record looks like this one from 5 October 2026, with readings that move around instead of climbing every day:

{
  "queryId": "7394231615091654662",
  "queryText": "Sourdough cookies",
  "popularity": 78,
  "searchVolume": 455736,
  "videoCount": 888,
  "trend7d": [285485, 275182, 288286, 172851, 284985, 117335, 455736],
  "trendScore7d": [74, 74, 74, 71, 74, 70, 78],
  "categoryL1": "Gourmet",
  "categoryL2": "Food Tutorials",
  "categoryL3": "Baking Tutorials",
  "isFavorited": false,
  "channel": "search",
  "tab": "SEARCH"
}

The content gap guide explains what TikTok counts as a match.

Field reference

Keep an absent field absent when you store these. A videoCount of zero and a videoCount that never arrived are different states, and collapsing them ruins any content-gap filter you build on top.

Follow-up tools

Each of these is one demand_topic call: pass the queryId from a topic record as id, plus a method.

The same tool also has related_queries, guidance (TikTok's content suggestions) and videos methods. The useful pattern is one broad demand_find_topics call, then follow-ups on the two or three topics that survive your filter. There is no point calling demand_topic method audience on all ten results, because you will drop most of them on volume alone.

Pagination, credits and limits

demand_find_topics takes limit from 1 to 20. The response returns hasMore and next, an opaque cursor; pass next back unchanged as cursor for the following page. The v2 search_topics call in the example returns nextOffset instead, which you pass back as offset: there it came back with nextOffset: 10. Either way, do not compute the step yourself from page numbers, because the server decides it.

Each lookup costs 1 credit and failed lookups are free. Free gives 50 credits once, 60 requests a minute and 2 at once; paid plans from $34.99 a month billed yearly start at 500 credits a month and 120 a minute (pricing).

Two practical notes. Keep the key out of source control and read it from the environment, as the example does. And if a call comes back with an upstream error, wait and retry once before failing the job: these are live TikTok reads, and a transient failure on one topic is not a reason to abandon the batch.

If you would rather ask questions than write a client, the desktop guide shows the same data through ChatGPT or Claude with no code at all.

What if you only need the answer?

If you only need the answer, not a pipeline, paste this into any assistant with TokConnect connected:

How we got these numbers

The outputs on this page come from TokConnect: "sourdough starter" (limit 5) on 19 September 2026 for the JSON record, and "sourdough starter" (limit 10) and "sourdough" (limit 20) on 5 October 2026 for the printed list, the filter results and the charts. searchVolume is TikTok's global search popularity for the last 7 days; searches per video is our division of searchVolume by videoCount. The Python examples were run against the server on 19 September 2026; the 5 October pulls used our own MCP research client with the same arguments. The tool count comes from TokConnect's live tool list, version 3.1.0.

TikTok's developer site returned errors to our automated checks on 5 October 2026, so the descriptions of the Research and Business APIs keep their 19 September reading; we found no announcement of a search-demand API in the meantime. Data: 5 October pulls, 19 September pull, tool catalog.