The DataForSEO API: SEO data by the cent

What does a real research question cost?

Receipts, not estimates: the keyword targets for this entire series — including this post — came from five DataForSEO Labs calls made while planning it. Roughly 180 keyword rows with volume, difficulty, and CPC, for a grand total of about eight cents:

What those cents bought, decision-wise: they revealed that "google search console api" carries the volume, "ga4 data api" is nearly uncontested, and a planned post angle had too little search demand to lead with — which reordered the series before a word was written. That's the entire pitch for paying per call: the data arrives exactly when a decision needs it, and costs nothing the rest of the month.

How do you set up DataForSEO, step by step?

1. Register at dataforseo.com — new accounts get a small trial credit, enough to try everything in this post many times over.
2. Get the API password from Dashboard → API Access. This is the classic first stumble: the API uses HTTP Basic auth with your account email as login and a separate API password — not the password you log into the website with.
3. Store the two halves as two env vars (LOGIN and PASSWORD) rather than one pre-encoded base64 blob. When a request 401s, "which half is wrong?" is the only useful question, and a blob hides it. Base64 is encoding, not encryption — it protects nothing anyway.
4. Verify with the free account endpoint before spending anything:

# Free endpoint — proves the credentials and shows the balance
curl -su "$DATAFORSEO_LOGIN:$DATAFORSEO_PASSWORD" \
  https://api.dataforseo.com/v3/appendix/user_data

# → "money": { "balance": 0.85, … }

Same reflex as the check scripts in the Google service account guide: prove the credential with a free probe, then configure from what the API answers.

Which endpoints give the most value per cent?

The API surface is enormous — hundreds of endpoints across SERPs, backlinks, on-page audits, merchant data. For content research, one family does most of the work: DataForSEO Labs. Seed it with a phrase and get back suggestions with volume, difficulty, and CPC in a single call:

POST /v3/dataforseo_labs/google/keyword_suggestions/live

[{
  "keyword": "server side tagging",
  "location_code": 2840,          // United States
  "language_code": "en",
  "limit": 60,
  "include_seed_keyword": true
}]

// each row: keyword, search_volume, cpc, keyword_difficulty
// each response: a "cost" field — the invoice travels with the data

• Set the limit deliberately — cost scales with rows returned. Sixty rows answered every question above; the default is far higher.
• Live endpoints take one task per request. Batch several tasks into one POST and the extras return "You can set only one task at a time" — the queued (non-live) endpoints are the ones that accept task arrays, trading latency for batching.
• Read the seed's own row first (include_seed_keyword: true) — volume and difficulty for the exact head term is usually the decision.

How do you keep the balance from evaporating?

Per-call pricing has a failure mode subscriptions don't: a loop with a bug spends real money at machine speed. The discipline that kept this series' research at eight cents:

1. Set a budget per task before the first call — this series ran under an explicit ten-cent cap.
2. Check the balance (free) before and after a research session; the delta is the session's true cost.
3. Log the cost field from every response next to the question it answered — the invoice line-items itself.
4. Keep the prepaid balance small. It doubles as a hard stop: an account holding $20 cannot have a $2,000 accident.

When does Semrush earn its subscription?

The honest comparison isn't data quality — everyone models volume from similar signals. It's what you're paying for: raw data versus finished workflows. As of writing, entry plans on the Semrush-class platforms run north of $100 a month, and here is what that buys over the per-call API:

The two models also compose: platforms for the humans who live in SEO all day, the API for pipelines — where keyword data joins the Search Console archive and flows into reports sized for agents. Same discipline as every other API in this stack: pull deliberately, log the cost, cache the answer.

Questions