Keyword research with Google autocomplete: the method and its limits

Autocomplete shows completions from real searches. Google weights them by popularity and freshness. It filters for policy. It varies by language, location and your own history.

It is a list of what people type, not volumes. Treat it as demand signals, not forecasts. Use it to find wording, modifiers and gaps.

On a small site you must pick your battles. Autocomplete tells you how searchers phrase the job. That makes your titles, H1s and sections match intent faster.

You want neutral results. Your history biases suggestions. So prepare a clean view before you collect anything.

1. 1

   Use a fresh browser profile or a private window

   Log out of Google. Disable sync. Close extensions that rewrite results.

2. 2

   Pick the right Google domain and language

   Use the country domain that matches your market, for example google.co.uk for the UK. In Search settings set Language and Region to your target audience.

3. 3

   Set your location

   Allow browser location if local intent matters, or use a reliable location switch in Search settings. For a SaaS with global reach, keep it broad.

4. 4

   Repeat for each market

   If you sell in English and German, repeat the process in each language and country combination.

You can also use a visualiser. AnswerThePublic shows autocomplete questions and prepositions. It is useful for a first pass. Manually verify any idea in Google after.

Start from a seed. Type your category name or the job your product does. Then fan out, one character at a time. Capture the list before it changes.

1. 1

   Choose a solid seed

   Pick a term that searchers would use, for example “note taking app”, “kanban board”, or “stripe integration”.

2. 2

   Add a space then a letter

   Type “note taking app a”, “note taking app b”, through to “note taking app z”. Copy the suggestions each time.

3. 3

   Try a leading letter

   Type “a note taking app”, “best note taking app”, “cheap note taking app”. Leading words change intent a lot.

4. 4

   Use a wildcard

   Insert a space in the middle, for example “note taking for ” then a letter. It surfaces audience variants like “for students” or “for designers”.

Work in tight bursts. Suggestions shift with freshness. Take what you see in one sitting per market. Name your file with date and market, for example “uk-en-2026-09-14”.

Modifiers turn a generic idea into a plan. Use the ones that map to the intents you can serve today. Here are the most useful for product sites.

• vs: “notion vs obsidian”, “yourproduct vs x”. Commercial investigation. Good for /compare/x-vs-y pages.
• alternatives: “x alternatives”, “alternatives to x”. Commercial investigation. Good for /alternatives/x-alternatives.
• pricing: “x pricing”, “x cost”. Transactional. Good for /pricing and rival pricing explains.
• how to: “how to do z”, “how to x with y”. Informational. Good for guides and use cases.
• for: “x for y”. Informational or commercial. Good for audience and industry pages.

Also mine prepositions and attributes. Examples: “with”, “without”, “on”, “in”, “template”, “examples”, “checklist”. Each suggests a section or a page depending on scope.

Look at the results page to confirm intent. Guides and videos mean informational. Product pages and shopping mean transactional. Lists and comparisons mean commercial investigation.

You need one table with your raw suggestions, a clean column, an intent column, and a notes column. Keep one sheet per market. Add a tab for rivals later.

1. 1

   Paste your raw list

   Collect suggestions per seed until they start to repeat. Use plain text. One query per row. This is typical, not measured.

2. 2

   Normalise and deduplicate

   Lowercase, trim spaces, unify spellings. Remove duplicates and obvious noise, for example typos that you would not target.

3. 3

   Classify intent

   Tag each as informational, navigational, commercial investigation or transactional. Use the results page layout as your guide.

4. 4

   Group close variants

   Cluster by meaning, not spelling. “x pricing” and “how much is x” live together. Plan one page per cluster unless intent splits.

A clean row might read: “kanban board for agencies”, intent commercial investigation, notes “/solutions/agencies or a section on /solutions” with the SERP showing list posts and product pages mixed.

Suggestions shift by market. Do not translate your English list and call it done. Build lists natively for each language and country you target.

• Use the local Google domain and set Search settings to the right Language and Region.
• Repeat the letter runs in the local language, for example “rechnungstool” in Germany, not “invoice tool”.
• Collect local modifiers, for example “preis”, “kosten”, “vergleich” in German, “tarifs”, “avis” in French.
• Map to local pages where it matters, for example localised /pricing and /compare pages.

Keep the language lists separate. Do not merge English and German into one cluster. Your URLs, titles and internal links will diverge by market.

Autocomplete does not show volume. You need another source to size clusters. Use the free options first.

• Bing Webmaster Tools keyword research shows volumes for free. Use it for a sense check by market.
• Google Ads Keyword Planner shows ranges to accounts without spend and figures to accounts with it. It groups close variants.
• Third-party tools estimate from Google Ads data, clickstream panels and models. Two tools will disagree. Treat both as approximations.

Sum volume at the cluster level. Plan for lower clicks when the SERP has a featured snippet or an AI Overview. The number on a tool is not a click promise.

Keyword difficulty is a tool estimate from the links of top results. It differs per tool and says nothing about intent fit. Open the SERP and judge fit yourself.

Map each cluster to one page type. Use your site structure, not just blog posts. For a SaaS, these families earn and compound over time.

• Category and “best x” pages: list the space and where you sit. Example: /guides/best-kanban-tools.
• Alternatives and “x vs y”: comparison content. Examples: /alternatives/trello-alternatives, /compare/trello-vs-yourproduct.
• Jobs to be done: “how to do z” guides. Example: /guides/how-to-set-up-kanban-swimlanes.
• Integrations: “x with y”. Example: /integrations/slack-kanban.
• Pricing and competitor pricing explainers: /pricing and posts like “notion pricing explained” if you sell against them.
• Your brand: make sure /, /pricing, /login, and /guides/getting-started match branded search.

1. 1

   Prototype the page outline from the cluster

   List the subtopics from suggestions as H2s. Remove fluff. Match the SERP’s depth.

2. 2

   Write the title and H1 to spec

   Keep titles near 60 characters. Include the main phrase. Write a clear H1 that mirrors it without stuffing.

3. 3

   Place internal links that help discovery

   Link from related pages and your homepage if it fits. Use descriptive anchor text that matches the intent.

4. 4

   Publish and check Search Console

   Watch the queries your new page starts earning. Expand sections that get impressions but thin clicks.

A fixed page looks like “/compare/trello-vs-yourproduct” with a direct title, a clear verdict, and sections that cover the common modifiers you saw in suggestions.

Do not build in a vacuum. Check where you already appear and where rivals earn. That makes your next page practical, not theoretical.

• Search Console’s query table shows where you already get impressions. Expand clusters where you sit in striking distance.
• site:rival.com searches reveal what Google holds of a rival. Their title tags show which queries they target today.
• A rival’s sitemap.xml lists their pages. It shows which comparison and solutions pages they invest in.

Pick the next page with three checks: clear intent fit, a realistic click share given the SERP features, and a path to link to it from pages you already have.

Autocomplete is personalised. It reflects your location and search history. Use a clean setup and repeat per market to offset this bias.

There are no volumes. Do not treat one suggestion as a market. Cluster by meaning, then size the cluster with external data if you must.

Freshness shifts lists. Trends and news will move suggestions by the week. Date your exports. Revisit quarterly, not daily.

Policy filters hide some terms. Absence in autocomplete does not mean zero demand. Cross check with People also ask and Related searches on the results page.