Autocomplete Topic Ideas: A Collection Method and Its Limits

Autocomplete suggestions are a free source of topic ideas, and a weak source of anything else. They show you how real searchers phrase a topic, but they do not rank by popularity, they vary from person to person and they do not tell you how many people are searching. If you write for a shop that sells used camera lenses, the useful output is a list of question and comparison patterns your readers type. The decision about which of those patterns deserves a page comes from other tools.

What autocomplete is

Predictions built from real searches

Google's help page on autocomplete (Opens in a new tab) says predictions reflect real searches that have been done on Google. The systems match queries to what you type, and they take into account the language, the location, trending interest and your past searches. The page adds that autocomplete can also suggest individual words and phrases based on real searches and word patterns across the web.

That gives the method its strength and its limit. The strength is that suggestions come from real behaviour. The limit is that two people typing the same letters may see different lists.

A prediction is not a verdict

The page notes that predictions are not assertions of facts or opinions, even though some may be perceived that way. Autocomplete does not stop you from typing and running a full search, and the page says it does not simply display the most frequently typed queries. For that reason it should not be compared to Google Trends, which the page describes as a research tool for search popularity over time.

For the used lens shop, the first suggestion that appears under "used 50mm lens" is not necessarily the top-searched phrase. Practical suggestion: record every suggestion as an idea candidate and ignore its position in the list.

Why two people see different lists

According to the page, a prediction can be unique to a particular location or time. The factors it names are in the table.

Trending interest

The page says current interest helps with events such as breaking news. It also says trending searches appear under the search bar on google.com, but the feature is not available in all languages or countries. For a camera lens topic, a product launch or a photography event can reshape the list for a while, so a dated record matters.

Personalization

Signed-in users may see a history of past searches and personalized predictions based on their activity on Google, and the page says users can manage those settings. Practical suggestion: if you do not want your own browsing history to shape the list, collect signed out, in a separate browser profile, and write that down. If the interface is in another language for you, labels may differ from the English names given.

Users can hide predictions

The page says users can turn off, remove or report certain predictions. That is one more reason two lists differ.

A collection protocol you can repeat

A list without its conditions cannot be reproduced three months later. Record the fields in this table for every session.

Stay manual

The method is manual observation. For bulk or automated querying, check the terms of the service you would be querying. Practical suggestion: collect by hand and stop when new rounds stop adding new wording.

Choosing seed phrases

Expansion quality depends on the seed. A random seed gives random suggestions.

Three sources

• Your own product names: the catalogue name may differ from the words shoppers use. Try both.
• Customer messages: pre-sale questions in emails and chats are the closest seeds to real wording.
• Search Console queries: phrases your site already appears for. Add parts to them to find neighbours.

Table: Seeds for a used lens shop

Keep the seed count small

Practical suggestion: three or four seeds per session. Each seed multiplies into letter, question and preposition expansions, so the record grows fast. An overlong seed list makes you log the same question again and again.

Seeds taken from friction points

A good seed can come from a spot where shoppers hesitate. For used lenses, hesitation sits around condition, compatibility and trust: what a cosmetic mark means, whether a lens fits a particular body, how returns work. Write each friction point as a short phrase and use it as a seed. The suggestions that come back show how shoppers word the worry, and that wording is better material for a guide than anything invented at a desk.

Seeds you should skip

Skip seeds that name brands you do not stock, products outside your catalogue and phrases so broad that the expansion will wander across the whole of photography. A narrow seed gives a narrow, usable list. A seed that is too broad produces rows that you will end up moving to the not-a-fit tab.

Expansion patterns

Letters

Add one letter after the phrase and record the suggestions: "used 50mm lens a", "used 50mm lens b". Each letter can open a different set of suggestions, though a set is not guaranteed for every letter. If your language uses accented letters, test them too. For languages with heavy suffixing, such as Turkish, see keyword research for suffixes and inflections.

Question words

Put how, why, when, which, where or can before or after the phrase: "how to check a used lens", "which used 50mm lens". These patterns can expose the questions behind a purchase, and they become guide headings and FAQ entries.

Prepositions and modifiers

Add for, with, under, near, vs. "used 50mm lens for portraits" or "used lens with adapter" reveal use cases. These fit category page copy and buying guides.

Comparison patterns

Words such as "or", "vs" and "versus" open comparison intent: "prime lens or zoom lens for travel". These produce decision articles.

Table: Expansion and output

Logging what did not appear

Keep a short log of seeds that returned nothing and of parts that produced no new suggestion. The log prevents the team from repeating dead ends and shows where another source is needed. When a seed returns nothing, note whether it was a short phrase, a long phrase or a phrase with a model number, since the cure differs: shorten the long ones, simplify the model number and try a different question word for the short ones.

When to stop

Use two stopping rules. First, stop a seed when several parts in a row bring nothing new. Second, stop adding seeds when the list is already bigger than your content plan can carry. The goal is a usable topic list, not the longest list.

What gets removed, and what a missing suggestion means

The page says automated systems try to block predictions that are violent, sexually explicit, hateful, disparaging or dangerous, including unconfirmed rumors after news events, and that enforcement teams remove violating predictions that slip through, along with closely related variations. It lists policies for general content, elections, health, named individuals and serious malevolent acts, and says the policies may be relaxed for artistic, educational, historical, documentary or scientific content, or content that aids understanding of current events and societal issues.

What it means for your list

A topic missing from the suggestions does not prove that nobody searches for it. A prediction may be withheld under policy. Practical suggestion: do not read the absence of a suggestion as an absence of demand.

When the list comes back empty

If a phrase returns no suggestions, change the wording first, then shorten the phrase, then try another question word. If it is still empty, look for the topic in Search Console queries and in customer messages. An empty list says only that this surface gave no data for that phrase.

Reading suggestions as language, not as numbers

The soundest use of a suggestion is as evidence of wording. If shoppers type "lens haze" in one place and "foggy lens" in another, both phrasings belong in the guide, one as the heading and the other as a natural variant in the text. Wording evidence is useful because it comes from real searches, and it needs no volume figure to be useful. Claims about popularity are a different matter and need a different source.

Suggestions are not volume

The autocomplete page says it should not be compared to Google Trends. The Trends data FAQ (Opens in a new tab) adds that Trends is not a scientific poll, that a spike does not mean a topic is winning and that its data should be treated as one data point among others. The same caution fits autocomplete: one suggestion is one data point.

Division of labour

• Autocomplete: question patterns and phrasing variety.
• Trends: the direction of interest over time.
• Search Console: what your own site really receives.

Table: Reading the three together

Example scenario: a used camera lens topic list

Example scenario: a shop that sells used lenses plans its blog. The session has six steps.

1. Choose seeds: used camera lens, used 50mm lens, lens fungus.
2. In a clean, signed-out browser profile, run letter and question expansions and log every suggestion exactly as shown.
3. Add preposition and comparison patterns: for, with, vs, or.
4. Label each suggestion by intent: information, decision, purchase.
5. Merge suggestions that ask the same thing in different words.
6. Move suggestions about products you do not sell to a "not a fit" tab.

The resulting table

These rows are illustrations. Collect your own list by hand.

A one hour session plan

Practical suggestion: put the collection in a time box.

1. First fifteen minutes: set conditions, write seeds, open the record table.
2. Next twenty minutes: letter and question expansions, logging text exactly as shown.
3. Next fifteen minutes: preposition and comparison patterns.
4. Last ten minutes: merge duplicates and add intent labels.

Verification is a separate session. Squeezing both into one makes it harder to judge the quality of an idea.

From suggestions to a topic list and a cluster

Cleaning the raw list

A raw list has three problems: the same question repeated in different words, suggestions outside what you sell and fragments that are not ideas. Use this vocabulary for labels.

• Information: how, why, what is, how much.
• Decision: vs, or, which, difference.
• Purchase: buy, price, where.
• Brand or place: names that are not yours, specific cities.
• Not a fit: things you do not sell.

Cleaning steps:

1. Merge rows that ask the same question and keep the wording a customer would use.
2. Move not-a-fit rows to a separate tab instead of deleting them.
3. Send brand and place rows to brand or local page plans.
4. Mark information and decision rows as article candidates.
5. Move purchase rows into category and product copy plans.

Attaching rows to pages

Short questions with short answers go to the FAQ block of an existing page. Decision questions get their own comparison article. Purchase patterns go into the category page. To see how long phrases fit a wider plan, read the long-tail keyword strategy guide.

Priority questions

Practical suggestion: score each row with three yes or no questions.

Rows with three yes answers go into the first wave, two yes answers into the second and one into a waiting list. No figures are needed, only three answers.

Verify before you write

A suggestion is an idea, not evidence. Before turning one into an article, look at the results page for the page types it returns, search Search Console for similar queries and check the related keyword in a tool such as Keyword Planner for a sense of size. If all three come back empty, move the idea to a low-priority list.

From suggestion to headline

A suggestion is how the reader types. A headline is a promise. Do not paste the suggestion in as the title. Write a natural sentence that names the question the page answers, and keep the wording of the suggestion for the body where it fits. Example scenario: "how to check a used lens" becomes "How to inspect a used lens before you buy", with a checklist under the headline. The phrase is not repeated mechanically across the text.

Keeping the list alive

A topic list ages. Lens models change, new mounts arrive and shoppers learn new words. Repeat the exercise on a calendar, not on impulse.

A simple rhythm

• After each new product category goes live, run a short session for that category.
• Once a quarter, repeat the full session with the same seeds and the same conditions, then compare the two records.
• Whenever Search Console shows a new cluster of queries, add one of them as a seed.

Comparing two records

Put the earlier record and the new one side by side. Suggestions that appear in both are stable phrasing. Suggestions that appear only in the new record may show a change in wording or interest, and those are the ones to check in Trends. Suggestions that vanished may have been removed under policy, may have faded or may reflect a change in the conditions of collection, which is why the conditions column matters.

Rules and checklist

• Rule 1: Do not treat suggestion order as importance.
• Rule 2: Record date, language, location and signed-in state for every list.
• Rule 3: Merge suggestions that ask the same question.
• Rule 4: Do not read a missing suggestion as missing demand.
• Rule 5: Support any popularity claim with Trends, Keyword Planner or Search Console.
• Rule 6: Remove suggestions for things you do not sell from the topic list.

Checklist:

• Seeds and conditions are in the record.
• Letter, question, preposition and comparison expansions were tried.
• Accented characters and variants were tested where the language needs them.
• Suggestions carry intent labels.
• Duplicates are merged.
• Results pages and Search Console were checked before writing.
• The date and conditions sit at the top of the list.

Pitfalls

• Reading suggestion order as popularity order.
• Treating a personalized list as the list everyone sees.
• Not recording the collection conditions.
• Deleting a topic because it did not appear as a suggestion.
• Assuming every letter yields suggestions.
• Turning a suggestion into an article without verification.
• Querying in bulk without checking the terms of service.

Each pitfall has a one-line cure in the template. A "signed-in state" column separates personalized lists. A "date" column keeps trending effects visible. A "verified" column marks the rows that passed the three checks. When you report to a manager or a client, label the document an idea list, not a demand report, and put the date and the conditions in its first line. Repeat the exercise at a later date and place the two records side by side. Suggestions that appear in only one of them point to a shift in interest or to a one-off event, and they deserve a closer look in Trends before anyone commissions an article.

Repeating the collection to keep the list current

A list collected once ages. Because current interest and location keep shaping suggestions, repeat the collection under the same conditions and set two records side by side. For a shop tied to seasons or product launches, do it before the busy period. A refresh does not mean starting over: add a new date column to the existing log and retry the same seed phrases under the same conditions.

Refresh triggers

Comparing two records

Put the old and new lists side by side and sort the rows into three groups: present in both, present only in the new list and present only in the old one. Rows only in the new list are candidates for a new need. Rows only in the old list do not prove the topic has closed, because a difference in location, sign-in state or news can explain them. Tie each difference to a condition change first, and only when the conditions match carry it into a content decision.

Decision rule: a difference enters the content plan only when the conditions are identical and it repeats on at least two collection dates. A one-off difference stays in the observation notes.

Example scenario: turning new suggestions into a plan

Example scenario: the used lens shop runs a second collection round before a photography fair. The log from round one holds "used 50mm lens test" and "used 50mm lens for portraits". Round two adds "used 50mm lens fungus check" and "used 50mm lens return policy". The browser profile, language, city and signed-out state match, so the two new rows count as real new wording.

1. Both new rows receive an information label.
2. The fungus-check row is matched with the shop's own inspection routine. The answer comes from the shop's checks, not from the suggestion.
3. The return-policy row is answered from the written policy and placed in the FAQ of the category page.
4. Both rows are searched in Search Console queries and in customer messages. If they appear there, their priority rises.

In this flow the suggestion only reminds you of the question; the source of the answer is the shop's own knowledge. When the suggestion list and the shop's facts disagree, the shop's facts win, because you answer for the accuracy of what the reader is told.

A handover note for the team

When someone else takes over the collection, a short handover note lets that person reproduce the result. The note sits next to the log and is updated at each refresh.

• Which seed phrases were used and why they were chosen.
• Which browser profile, language, location and sign-in state applied.
• Which expansions were tried and which returned nothing.
• Which rows were merged and which moved to the not-a-fit tab.
• The trigger and date for the next refresh.

Handover checklist:

• Does the note sit beside the log?
• Was the gap between two collections separated from a condition change?
• Were the new rows checked against Search Console and customer messages?
• Is the next refresh date on the calendar?

Frequently Asked Questions

How are autocomplete predictions generated?

Per Google's help page, predictions reflect real searches done on Google. The systems match queries to what you type and consider language, location, trending interest and your past searches.

Why do two people see different suggestions?

The page names location, language, trending interest and, for signed-in users, past searches. Users can also turn off, remove or report certain predictions.

Do suggestions show search volume?

No. The page says autocomplete does not simply display the most frequently typed queries and should not be compared to Google Trends.

Why does a topic return no suggestions?

The page lists policies under which some predictions are blocked or removed. A missing suggestion does not show that nobody searches for the topic.

How do letters and question words help?

Each added part opens another set of suggestions. Practical suggestion: try letters and question words in turn and log every suggestion exactly as shown.

Can I collect suggestions in bulk with a script?

The documentation has no statement on bulk collection. Check the terms of the service you would query. The method is manual observation.

What should I check before writing about a suggestion?

Look at the results page, search Search Console for similar queries and check the related keyword in a tool such as Keyword Planner for a sense of size.

Conclusion

Autocomplete gives you the phrasing and question patterns of real searchers without a subscription. The order of suggestions is not a popularity ranking, the list changes with person and place and policy limits hide some predictions. Record your conditions, label suggestions by intent and verify an idea elsewhere before you commit a page to it.

Once the used lens topic list is ready, the next job is writing articles that answer those questions with care and sources. If you want a partner for that step, have a look at our content writing service.