Rankings and SEO research

The Rankings workspace puts classic search position, AI visibility, keyword opportunity, and competitor evidence on one query record.

Know which source answered the question

Source	What it tells you	Important limitation
Google Search Console	Clicks, impressions, CTR, and average position Google observed for the connected property	Only queries with property data; reporting delay and Google retention apply
Bing Webmaster Tools	Bing query and position data for the connected site	Bing data, not Google rank
DataForSEO direct SERP check	The Google result Tailwin requested for a keyword, location, language, depth, and day	Metered point-in-time observation
DataForSEO Labs	Estimated volume, difficulty, domain footprint, related terms, pages, and backlinks	Vendor estimate; not the direct rank Tailwin observed
Signal	Whether AI engines named or cited the brand for the query	AI answers, not classic SERP position

Tailwin keeps these labels separate. Never describe Search Console impressions as search volume or Labs estimates as a rank check.

Add tracked keywords in bulk

1. Open Rankings.
2. Choose the intended site.
3. Paste one keyword or buyer question per line.
4. Submit the list.
5. Review the accepted, duplicate, invalid, and over-limit results.

Every active query up to the plan limit can participate in scheduled direct checks. The hard safety ceiling is 200 keywords per source sync, above the current paid plan limit of 100.

How direct Google checks control cost

Tailwin uses DataForSEO’s Standard Queue for organic results:

• once each week, every tracked keyword receives a top-100 discovery check;
• between discovery runs, only new keywords and keywords most recently found in the top 20 receive a top-20 daily check;
• volume, difficulty, and backlink summary calls run only with the weekly discovery cycle;
• same-day task tags and DataForSEO’s task ledger recover accepted work after a retry instead of purchasing it twice;
• posts are batched, result polling is concurrency-limited, and transient errors back off.

A value of 21 or 101 can mean “checked and not found within this run’s depth.” The stored payload includes the checked depth so the interface can distinguish that from “not checked.”

Run keyword and competitor research

Use the SEO workbench to:

1. seed a phrase and retrieve related terms or ideas;
2. inspect estimated volume, difficulty, CPC, and intent;
3. inspect a domain’s estimated organic footprint and ranked pages;
4. retrieve referring domains when the account includes Backlinks access;
5. add selected terms to the tracked query set.

Research results are cached per organization for 12 hours to reduce repeated vendor charges. If DataForSEO is unavailable, the MCP and UI return an explicit unavailability response; agent skills fall back to measured Search Console queries where appropriate and must not fabricate estimates.

Use the white-label MCP

Agency and operator integrations can create twm_ MCP keys in Settings and call /api/mcp. The tools expose the same organization-scoped facts and honest fallback rules as the UI. Keys are secrets: copy a new value once, store it in a secret manager, and revoke it when no longer needed.

The audit_url tool scores one public page for citability, AI readability, robots access and structured data. The structured data section lists JSON-LD types, the rich-result state for each entity (eligible, missing required, or deprecated), blocking findings, and the rules version used. It makes no AI calls.

Audit tools and cost confirmation

Start with list_sites and use a returned site ID. The key fixes the organization scope; agents cannot supply another organization or act as an individual user's saved chat preferences.

Tool	What it does	Cost class
inspect_schema	Inspect an owned page and retain a schema snapshot	Free guarded fetch and computation
generate_json_ld	Prepare a facts-only JSON-LD draft from the page and verified operator facts	Free guarded fetch and computation; no LLM
run_pagespeed	Request mobile or desktop Lighthouse lab and CrUX field evidence	Google quota; explicit confirmation
list_audit_snapshots	Read up to 50 retained snapshots for a site/page	Free stored read
compare_audits	Compare chronological, compatible snapshots	Free stored read
get_fix_brief	Export deterministic fixes, captured targets and acceptance checks	Free stored read
rerun_audit	Rerun a selected baseline's page and scope, then compare	Fresh fetch; speed reruns consume Google quota
analyze_competitor_page	Analyze one competitor page plus page-scoped keyword estimates	Paid DataForSEO Labs research; explicit confirmation

For PageSpeed, specify strategy: "mobile" or "desktop" and set confirmMetered: true only after the operator authorizes quota use. Keep Lighthouse lab metrics separate from CrUX field observations, including whether field coverage refers to a page or its origin. Missing data is unavailable, not zero. A valid speed receipt can be reused for 12 hours; new requests share the workspace's 50-per-day limit. A speed rerun bypasses that cache, preserves quota and in-flight limits, and compares against the exact selected baseline.

History reads need real snapshot IDs from list_audit_snapshots. Comparisons preserve source dates, page/category scope, rules and scorer versions. A failed or unavailable category does not prove its earlier findings resolved. A fix brief's verified-property link opens Search Console inspection; it does not request indexing.

Single-page competitor analysis accepts siteId, one public HTTPS url, optional locationCode and languageCode, and confirmMetered: true. Defaults are United States (2840) and English (en). New analysis requests authorize up to $0.02 of reviewed Labs cost and are limited to 20 per workspace per day; valid cached results can be reused for 12 hours. Static page facts and vendor keyword estimates remain separate. The tool checks robots access, does not render JavaScript or crawl links, and does not return the retained private page text. An uncertain charge needs review rather than automatic retry. Creating an original Content draft is a separate, explicitly approved action.

Legacy integration change: research_keywords, get_domain_overview, get_ranked_keywords and get_backlinks_overview now also require confirmMetered: true before provider work. Update older agent calls to obtain authorization and include that field. Their multi-request costs differ from the single-page estimate. add_rank_tracking_keywords remains a free database mutation; it adds the approved terms under plan limits without immediately dispatching a paid scan.

The in-app /docs/skills guides provide JSON tools/call parameter examples for each workflow. Replace their example site/snapshot IDs and page URLs with real scoped values. Schema generation uses verified facts only; review missing requirements before installing a draft. No tool guarantees rich results, indexing or ranking improvement. These guides document implemented functionality, not a live-provider canary or proof that credentials are configured.

When to use your own data collection

Tailwin should crawl pages you own or are authorized to audit for content, technical, and change data. It should not operate an unapproved first-party Google results scraper. Search result collection is legally and operationally different from crawling a customer’s own site: blocking, localization, anti-bot infrastructure, and terms risk make it an expensive commodity rail. The product rents the result index and owns the retained history.

The site page includes an owned-corpus crawler for customer-authorized pages, not a Google SERP clone. Its findings include blocking structured data problems for each crawled page.

YouTube search positions

In Rankings, YouTube search ranks measures a target video for a search phrase. Add its 11-character video ID, choose the site and expand Location and search depth when needed. Defaults are United States (location code 2840), English, desktop and 20 result blocks. You can request mobile or 100 blocks. Location/language codes must be supported by DataForSEO.

Select the metered authorization checkbox to queue a paid check. Deeper checks can consume more vendor credit. The background worker processes saved checks; Refresh rank history shows progress. Identical checks for the same site and UTC day reuse a receipt. The daily workspace cap is the smaller of 20 checks and the plan's per-site keyword allowance. This control limits vendor usage; it does not establish subscription billing readiness.

The table shows absolute video position, location, language, device, requested block depth and deepest returned block. IDs reveals the run/vendor references, video-group position and block position. Not observed means the target was absent from this returned sample. It does not mean the video cannot rank elsewhere. Observation timestamps come from the provider; pending rows show request time.

An uncertain task is recovered without purchasing another check. After 24 hours, unresolved work stops automatic recovery and keeps its diagnostic receipt. Charges shown are vendor task receipts, not invoice reconciliation. Latest 100 checks are shown. Channel audience and retention reports remain under Reports.

Discover keywords with Google autocomplete

In Rankings, owners and admins can use Keyword discovery to enter a niche phrase, choose a market and language, and collect suggestions for the phrase plus each letter A–Z. Confirm the metered lookup before running it: a new run makes up to 27 DataForSEO autocomplete requests. Tailwin reuses the same workspace's result for 12 hours and allows at most 10 new runs per rolling 24 hours.

Filter by keyword text or source query, use Select visible, and choose Track selected to add terms to your site's existing keyword capacity. Selection stays selected when you change filters; the selected count covers every filter. Export CSV downloads the currently visible suggestions with their source queries and collection context. A site is required for tracking, but discovery and export work without one.

Autocomplete suggestions are ideas, not search-volume or difficulty estimates. The collection shows its original seed, market, language and timestamp. Failed letters appear as unavailable while successful letters remain usable. Partial and failed results are also retained for 12 hours to prevent repeated charges. Changing the seed or locale starts a separate lookup. These lookups do not change the scheduled organic rank-check cadence.