tokconnectStart with the free manual pass
Use TikTok’s Creator Search Insights in the app to enter a phrase, open the visible topic cards, and record the exact topic labels that match the customer’s job. Then type the same phrase into TikTok search and preserve the autocomplete suggestions that change the task. These two passes cost no TokConnect credits and keep a strategist from treating the first returned phrase as the answer.
For a pet-care brand, the job can be “show a safe at-home bathing routine,” not the broad phrase “pet grooming.” Write the job before you collect numbers. A topic card that names birthdays, salon services, or an unrelated animal-health concern can still score highly while missing the planned video.
Manual pass: search “pet grooming at home” in Creator Search Insights and TikTok search. Save the visible topic label, the exact autocomplete phrase, and a screen link or note. Keep only phrases that describe an at-home routine. Reject phrases that describe celebrations, local services, or unrelated health questions.
Record the rejected phrases too. They show why the team chose a narrower route, and they stop the next person from reopening the same broad result because its score looks attractive. A free manual pass does not replace the MCP records. It gives the records a decision rule.
Retrieve topic records, then filter them
After the manual pass, a connected MCP client can call keyword_research with keyword: "pet grooming at home", language: "en", expand: 3, and videos: false. The saved 18 September 2026 result returned 20 matches.
The tool selected “pet birthday prep at home” as its seed. That record returned popularity 99 and a search-volume estimate of 119,535,102. Its label fails the at-home grooming rule, so the brief rejects it. A score does not override the job definition.
| Returned topic | Popularity | Search estimate | Brief decision |
|---|---|---|---|
| pet birthday prep at home | 99 | 119,535,102 | Reject: celebration preparation |
| pet spa grooming and wellness | 87 | 4,814,381 | Hold: mixes home and service language |
| pet bathing and grooming routines | 87 | 4,785,034 | Select for source review |
The field is a returned estimate. This result does not establish a geography, a monthly unit, buyers, or a total TikTok supply count. Read the metric guide before comparing topic records across a plan.
Score relevance before metrics
Use a simple written screen before you compare topic fields. For the pet-bathing job, the record must name a home routine, a bathing or grooming action, and a plausible viewer problem. Give each condition a yes or no. The selected record passes two strong wording tests: “pet bathing” names the action and “routines” fits the requested format. “Pet birthday prep at home” includes the location phrase but fails the action test. Its high returned estimate changes neither result.
| Screen | Pet bathing and grooming routines | Pet birthday prep at home | Decision |
|---|---|---|---|
| Names the customer task | Yes | No | Keep the bathing record |
| Fits a repeatable video format | Yes, a routine | Only for a celebration task | Keep the bathing record |
| Needs a claim outside your review | Possibly, for animal-care advice | Possibly | Require qualified claim review |
| Returned topic field | Popularity 87; estimate 4,785,034 | Popularity 99; estimate 119,535,102 | Use after the relevance screen |
Use the fields to choose where to look next only after that screen. A large estimate can justify inspecting a matching phrase. It cannot prove that a buyer will prefer a video, that a claim is safe, or that a video will rank. A small but exact match can be the stronger production choice when the job is narrow.
Select the matching record before expansion
keyword_research expands its top match. That behavior can produce related records from an irrelevant seed. The saved run’s automatic related list includes clothing and party topics, which do not belong in this brief. The researcher must select a matching record and run a follow-up with its queryId.
An 18 September follow-up used queryId 7668815927365861397 for “pet bathing and grooming routines” with related_topics, limit 3. It returned “bath and shower routines,” “order bath and body works,” and “Dog vomiting.” The first might inform a routine format; the second and third fail the pet-grooming job. The output demonstrates why a human filter belongs between retrieval and a content recommendation.
Turn the record into a video review
Retrieve five related or ordinary search results for the selected wording. Make a sheet with one row per returned video. Record the URL or ID, account, caption, obvious opening from watching the video, proof or demonstration step, disclosure or safety statement, and comments chosen for a later sample. The saved keyword record does not supply those production observations. Someone must open the videos.
Use the sheet to decide between formats. If three relevant videos lead with a before-and-after visual, that is an observed pattern in the five-video sample. If the planned product needs handling instructions, test an opening that names the bathing problem and then shows one verified step. Do not report “TikTok viewers prefer before-and-after videos” from five records. The deliverable is a testable script choice, not a universal audience rule.
Review five videos connected to query ID 7668815927365861397. For each, record the first visible claim, one demonstrated action, and the source URL. Select one pattern only if it appears in at least two reviewed videos. Draft a 20-second at-home bathing routine that names the product category, shows one verified handling step, and sends every animal-health claim for qualified review.
Use the comment analysis guide to preserve the source video, coding rule and denominator.
Use the pet-grooming brief example
| Brief field | Completed entry |
|---|---|
| Customer job | Show an at-home pet-bathing routine |
| Seed phrase | pet grooming at home |
| Selected record | pet bathing and grooming routines, queryId 7668815927365861397 |
| Rejected record | pet birthday prep at home, despite popularity 99 |
| Next evidence | Five related or ordinary video results; one bounded comment page |
| Proposed test | Open with the bathing problem, then show one handling step |
| Safety boundary | Do not make animal-care claims without qualified review |
Download the filled example brief. It records the retrieved evidence and an editorial selection for review, not a completed production decision. Your team can replace the selected source record after it reviews returned videos.
Stop or expand with a rule
Stop after the selected topic, five video rows, and one comment page when the team has a plausible script test and no unresolved claim blocks production. Expand when a source video shifts the task, when the returned phrase describes salon service instead of home care, or when comments ask a product question that the script cannot answer. Expansion has a purpose: another topic page, another five videos, or replies for one comment. “Look for more ideas” is not a purpose.
Keep manual research separate from retrieved records. The Creator Search Insights app and TikTok autocomplete can supply phrases a researcher saw, but they need a note or screenshot and cannot be merged with MCP values as if both came from one call. The completed brief should contain the input, selection rule, selected and rejected labels, downstream source IDs, proposed test, reviewer, and a date to revisit the choice.