How to Get Keyword Ideas & SERP Data With No API Key in 2026

Keyword research tools like DataForSEO, Ahrefs and SEMrush gate the core “keyword ideas + SERP + questions” job behind API keys, seat licenses and monthly subscriptions. For a lot of workflows — a content brief, a programmatic-SEO run, an AI agent that needs to research a niche mid-task — you don’t need volume estimates or difficulty scores; you need the discovery layer: what people actually search, what ranks right now, and what questions they ask. This guide covers how to pull all of that from public search sources into one structured record per seed keyword, with no API key.

What you get per keyword

Each seed keyword becomes exactly one dataset row, combining four data sources:

• autocompleteSuggestions — an array of Google autocomplete keyword ideas: the real queries users type, the gold standard for long-tail discovery.
• serpResults — the live organic SERP as an array of objects, each with rank, title, url, snippet and source — a lightweight built-in SERP API.
• peopleAlsoAsk — an array of People Also Ask questions, ready to structure FAQ sections and topic clusters.
• relatedKeywords — related searches for topic expansion, turning one seed into a content map.
• keyword / region — the seed and the region/language used.
• suggestionCount / serpCount — counts of ideas and SERP results returned.
• fetchedAt — ISO 8601 run timestamp.
• error — present only if a keyword failed entirely; the row is still saved.

One important honesty note up front, straight from the actor’s design: this returns keyword ideas, the SERP, and questions — the discovery side. It does not estimate monthly search volume or keyword difficulty. If your workflow depends on volume numbers, this is a complement to that data, not a replacement for it.

Where each field comes from

The actor doesn’t scrape a single engine — it pulls each part of the row from the source that does that job best, then merges:

• Keyword ideas — Google autocomplete (the Suggest JSON endpoint), the same suggestion engine behind Google’s search box.
• SERP — DuckDuckGo HTML organic results.
• People Also Ask — question blocks parsed from the Bing SERP (best-effort).
• Related keywords — Bing related searches plus DuckDuckGo related, merged and deduped.

Every source is best-effort and independently guarded: if one returns nothing, the others still fill the row, and the actor never crashes on a single keyword. That resilience matters at batch scale — one flaky source doesn’t sink the run.

Idea expansion has three modes:

• standard (default) — seed plus question prefixes and modifiers, yielding 100+ ideas.
• deep — also appends a–z suffixes, answer-the-public style, for maximum long-tail coverage.
• off — seed suggestions only.

A single-keyword full-research input:

{
  "keywords": ["mcp server"],
  "maxSerpResults": 10,
  "expansion": "standard",
  "region": "us-en"
}

A batch run with deep expansion in a German region:

{
  "keywords": ["kaffeevollautomat", "espressomaschine", "milchaufschäumer"],
  "expansion": "deep",
  "maxSerpResults": 15,
  "region": "de-de",
  "concurrency": 5
}

And ideas-plus-questions only, no SERP, for content briefs:

{
  "keyword": "ai agent framework",
  "includeSerp": false,
  "includePaa": true,
  "includeRelated": true,
  "expansion": "standard"
}

The region field (us-en, uk-en, de-de, fr-fr, es-es…) controls both the autocomplete language and the SERP region. maxSerpResults (1–30) caps organic results per keyword, and concurrency (1–10) sets how many keywords run in parallel.

▶ Run the SERP & Keyword Research actor on Apify — one seed in, 100+ keyword ideas plus the live SERP, People Also Ask and related searches out. No API key, no seat license. Export to JSON, CSV or Excel — or call it as an MCP tool.

The access reality: keyless, but rate-limited

There’s no API key to manage and no account to create on DataForSEO, Ahrefs, SEMrush, Moz or SerpAPI — the actor queries public suggestion and organic-result endpoints directly. The friction that remains is per-IP rate limiting, since you’re hitting Google Suggest, DuckDuckGo and Bing at volume:

• Proxy is on by default. useApifyProxy routes through Apify’s datacenter proxy to spread requests across IPs. For heavier batches, override with proxyGroups: ["RESIDENTIAL"].
• Concurrency is your throttle. concurrency (1–10) balances speed against rate-limit risk — higher finishes faster but hammers the sources harder.
• PAA is best-effort. People Also Ask blocks aren’t always surfaced; when none appear, peopleAlsoAsk returns an empty array rather than an error.

Because every source is guarded independently, a partial result is normal and expected: you might get autocomplete and SERP but an empty PAA list on a given keyword. That’s a design choice, not a failure — the row still saves.

Example output

A single keyword row, trimmed:

{
  "keyword": "mcp server",
  "region": "us-en",
  "autocompleteSuggestions": [
    "mcp server list",
    "mcp server python",
    "mcp server example",
    "mcp server claude desktop"
  ],
  "serpResults": [
    {
      "rank": 1,
      "title": "Model Context Protocol — Example Servers",
      "url": "https://modelcontextprotocol.io/examples",
      "snippet": "A collection of reference MCP server implementations...",
      "source": "modelcontextprotocol.io"
    }
  ],
  "peopleAlsoAsk": ["What is an MCP server?", "How do I build an MCP server?"],
  "relatedKeywords": ["mcp client", "mcp tools", "model context protocol"],
  "suggestionCount": 4,
  "serpCount": 10,
  "fetchedAt": "2026-07-06T12:00:00.000Z"
}

serpResults is an array of objects (not a flat string), so if you export to CSV, plan to parse or flatten it; JSON keeps the nesting intact.

Use cases

• SEO keyword research — turn one seed into hundreds of long-tail ideas in seconds via autocomplete plus expansion.
• Content briefs & FAQ pages — mine peopleAlsoAsk and relatedKeywords for the structure and sub-headings of a piece before you write it.
• Topic clustering — expand a niche into a full pillar-and-cluster map from related searches and idea expansion.
• Rank tracking & SERP analysis — capture who ranks for a term today (serpResults with rank, url, source) and re-run to watch positions move.
• AI agents & automation — a “research this keyword” tool an agent calls mid-task, paying per result, then reasons over the ideas and SERP.
• Programmatic SEO — feed thousands of seeds (batch up to 100 per run) and build data-driven landing pages at scale.

The AI-agent / MCP angle

This is designed to be callable by an AI agent, not just a human in a dashboard. The pay-per-result pricing is what makes that clean: an agent researches a keyword, pays per row returned, and reasons over the output — no seat license to provision, no key to inject. Wire it into Claude, Cursor or any MCP client as a keyless SEO tool and the agent does live keyword research and SERP analysis autonomously. For a human, the same run exports to JSON, CSV or Excel or reads live via the Apify API.

Build it yourself vs. use the managed actor

You could stitch together Google Suggest, a DuckDuckGo HTML parser and Bing scraping yourself. What you’d own:

• Four sources, four formats — the Suggest JSON, DuckDuckGo’s HTML, Bing’s PAA blocks and Bing/DDG related searches each parse differently.
• The merge and dedup — combining ideas and related searches from multiple sources without duplicates.
• Per-source resilience — guarding each source so one flaky endpoint doesn’t crash a 100-keyword batch, plus the proxy rotation to survive rate limits.
• Expansion logic — the question-prefix, modifier and a–z suffix generation for standard and deep modes.

Pricing is pay-per-result: one charge per keyword that returns data; keywords that return nothing are saved but not charged. No monthly fee, no seat license, no idle cost. Against a DataForSEO/Ahrefs/SEMrush subscription, that’s the difference between a fixed monthly bill and cents-per-keyword on demand — for the discovery portion of the job.

Common pitfalls

Specific gotchas when working with this data:

• No volume or difficulty, by design. Don’t build a workflow that expects search-volume or KD numbers from this actor — it’s the ideas/SERP/questions layer. Pair it with a volume source if you need those metrics.
• serpResults and other arrays are nested. They’re arrays of objects, not flat strings. CSV exports flatten awkwardly; use JSON when you need the structure, or flatten deliberately in your pipeline.
• PAA and related can be empty. Best-effort sources return empty arrays when nothing is surfaced. Treat an empty peopleAlsoAsk as “none found for this query,” not as a bug.
• Region controls both language and SERP. Setting region: de-de shifts autocomplete language and SERP region together. Don’t expect to mix an English SERP with German suggestions in one call.
• The SERP is DuckDuckGo, not Google. serpResults come from DuckDuckGo HTML. It’s a solid organic snapshot, but if you need Google’s exact ranking order specifically, know the source. Rankings correlate but aren’t identical.
• error rows still count as rows. A keyword that fails entirely is saved with an error field so you can see what happened — check for it before treating every row as a clean result.

Wrapping up

For the discovery side of keyword research — real long-tail ideas, the live SERP, the questions people ask — you don’t need a subscription or an API key. If you’re building content briefs, running programmatic SEO, or wiring a keyless research tool into an AI agent, a managed actor gives you all four sources merged into one row per seed, resilient to any single source failing.

▶ Open the SERP & Keyword Research actor on Apify — keyword ideas, live SERP, People Also Ask and related searches per seed, batch up to 100. No API key, no seat license. Export to JSON, CSV or Excel. Start on Apify’s free monthly credit.