Nine sources, ranked by how well the keywords convert
Every keyword source is a guess about what people type. The difference is whose guess it is. A keyword tool infers demand from clickstream panels and models the gaps. Your support inbox contains the actual sentence a real buyer wrote at 11pm, when they were confused enough to ask a stranger.
So this ranking isn't by search volume, and it isn't by how many ideas each source spits out. It's by how likely the resulting page is to produce a lead — which is the only ordering a business should care about.
| Source | What it gives you | Why it sits there |
|---|---|---|
| 1. Sales calls and demo recordings | The exact phrasing prospects use for their problem, before your marketing teaches them your vocabulary | Highest intent of anything on this list, and already transcribed if you record calls |
| 2. Support tickets, WhatsApp and live chat | Post-purchase questions, objections, and the words people use when something has gone wrong | Any question you answer twice a week is a page you should have published |
| 3. Your site's internal search log | What visitors couldn't find on a site they had already chosen to visit | Demand plus context. These people are further down the funnel than anyone on Google |
| 4. Google Search Console | Every query you got an impression for in 16 months, with clicks and average position | Proven demand where Google already treats you as a candidate |
| 5. Autocomplete, People Also Ask, related searches | Query variants Google has seen often enough to suggest unprompted | Free, effectively infinite, and it is Google describing its own view of the topic |
| 6. Competitor SERPs and sitemaps | Their titles, H2s, FAQ blocks and public XML sitemap | You inherit their research. A ranked page has already passed the demand test |
| 7. Communities and reviews | Reddit and Quora threads, industry WhatsApp groups, one-star reviews of the product you replace | Objections and comparison queries you would never have invented |
| 8. Keyword Planner, Trends, Bing Webmaster Tools | Bucketed volume ranges, seasonality, regional splits, and Bing's own query data | Free but modelled. Size a decision with these; don't make one |
| 9. Free tiers of the paid suites | Capped exports of competitor keywords and backlinks | Genuinely useful, genuinely limited, and the least specific to your business |
Sources 1–3: the words your buyers already used
This is the part most keyword guides skip, because it doesn't produce a spreadsheet in ten minutes. It's still the best hour you'll spend all quarter.
Pull the last twenty sales calls and the last hundred support tickets. Read them with one question in mind: what did they call it? Not what your product page calls it — what they called it before they learned your vocabulary. Founders are consistently surprised here. You sell workforce management; they search staff attendance app.
Internal site search is the same trick with less reading. If your CMS logs it, you already own a list of queries from people who liked you enough to visit. If it doesn't, turn on site search tracking in GA4 this week. It costs nothing and it starts paying by the next report.
- Live chat and WhatsApp — shorter, blunter, and closer to how people actually type into a search box than anything they'd say on a call.
- Sales objections — "how is this different from X" is a comparison page waiting to be written, and it converts better than anything above it.
Source 4: the demand you're already half-getting
Search Console's Performance report is the best free keyword source on this list, and most sites open it twice a year. It holds 16 months of queries — every search where your site appeared, whether or not anyone clicked.
Three filters do most of the work.
- Average position 8–20, sorted by impressions. Google already associates these queries with you but hasn't committed. A rewritten page, a sharper title, or two internal links often moves them into the range where clicks happen.
- Queries with impressions and no clicks. Either your page answers a different question, or your title is losing the click to someone else's. Both are cheap to diagnose and cheap to fix.
- Queries ranking on a page you never wrote for them. That's a page worth building properly, and frequently a keyword cannibalisation knot to untangle first.
Sources 5–7: Google's suggestions, and other people's pages
Once you've exhausted your own data, you're looking for demand you haven't earned yet. Three free places give it to you.
Autocomplete, People Also Ask and related searches
Type a seed into Google and read the suggestions before you press enter. Then add each letter of the alphabet after it and read again — that's a few hundred real variants in ten minutes. Open the People Also Ask boxes and watch them multiply as you go. Scroll to the related searches at the foot of the page.
Do it in an incognito window with your region set to India, because autocomplete is personalised and location-dependent. Otherwise you're researching your own browsing history.
Read competitor SERPs, not competitor traffic
You don't need a subscription to see what a competitor targeted. Their title tags, H1, H2 order and FAQ blocks tell you exactly. So does their XML sitemap, which is public by design — try /sitemap.xml, or read their robots.txt for the path.
A competitor's blog sitemap is a plain list of the topics somebody decided were worth paying a writer for. Whether it worked is a separate question, but it's free research either way.
Communities, reviews and the questions people ask strangers
Reddit and Quora threads now rank for a lot of Indian commercial queries, which is Google telling you it considers those threads a decent answer. Read them for phrasing, not for volume. And the one-star reviews of the product you're replacing are the richest source of comparison queries anywhere — pre-sorted by how annoyed the customer was.
Hinglish, transliteration and the queries tools underreport
Indian search is multilingual in a way most keyword databases handle badly. The same buyer might search a clean English phrase on Monday, a Hinglish version on Wednesday and a transliterated Hindi one on their phone that evening. Your tool will probably report the first and shrug at the rest.
You find those variants by going where they're recorded rather than where they're modelled.
- Autocomplete with the region set to India — set location, not just language. The suggestions genuinely differ.
- Google Trends region comparison — no absolute volume, but it will show you which phrasing is rising and which states search it.
- YouTube search — transliterated and regional queries surface there before they show up in any keyword tool.
- Your own WhatsApp and support threads — the most honest record in existence of how your customers phrase things.
- Bing Webmaster Tools — a different index with its own free keyword data. Smaller market, real numbers.
Sources 8–9: free tiers, and exactly where they stop
Two things are true at once. The free tiers are more useful than most people assume, and they cut off precisely where the interesting work starts.
- Google Keyword Planner — free with a Google Ads account, but volumes show as wide buckets unless the account is actively spending. Fine for relative sizing, useless for precision.
- Google Trends — relative interest over time, never absolute volume. Best for seasonality and for settling an argument between two phrasings.
- Bing Webmaster Tools — free keyword research from Bing's own query data, no ad spend required.
- Ahrefs Webmaster Tools — free for sites you can verify. Shows your backlinks and your ranking keywords; not your competitors'.
- Free tiers of the big suites — typically a few searches a day. Enough to check a hunch, not enough to plan a quarter.
You have 200 ideas. That still isn't a plan
A list is raw material. The next moves are grouping by intent, checking each group against the live SERP, and assigning exactly one primary keyword per URL — which is the whole of keyword research done properly.
Most small sites also need far fewer keywords than they collect. Twenty-five well-chosen ones, each with a page that genuinely answers the query, will out-earn a 4,000-row export nobody opens twice. Research volume was never the constraint. Publishing is.
If you'd rather hand the exercise over, that's what our keyword research work covers. Read your own support tickets first regardless — nobody can do it as well as the person who answered them.