A keyword is a question with an instruction attached
Most keyword advice treats a keyword as a string of words with a number next to it. That's the part tools can measure, so it's the part people optimise. It's also the least useful part.
What a keyword actually carries is an unspoken instruction about what the searcher expects to find. "Seo agency pricing" expects a table of numbers. "Best seo agency" expects a list. "How does seo work" expects an explanation. "Last agency" expects one specific website. Same three-word shape, four completely different pages required.
Google has already made that ruling for every keyword you're considering, and it publishes the ruling in plain sight — as the ten results currently on page one. Every page you build is a bet that you can satisfy the intent better than the incumbents. If your page format is wrong, you're not competing badly. You're not competing at all.
The four intents, and the page each one demands
Every commercial keyword set sorts into four buckets. The bucket decides the page type, the length, the tone, and whether the page belongs on your blog or in your navigation.
| Intent | The searcher wants | SERP tells you by | Page you build |
|---|---|---|---|
| Informational | To understand something. No purchase in mind yet. | Featured snippet, People Also Ask, videos, mostly guides and Wikipedia. | Article, explainer, FAQ. Depth wins. |
| Commercial | To compare before buying. Purchase intended, vendor undecided. | Listicles, review sites, comparison tables, some ads. | Comparison, alternatives page, honest pros-and-cons. |
| Transactional | To buy, book or sign up now. | Shopping results, heavy ad load, product and pricing pages. | Service page, product page, pricing page. |
| Navigational | To reach one specific site or brand. | One dominant result plus its sitelinks. | Nothing, unless it's your brand. Don't chase these. |
Read the SERP before you commit a page
This is the skill that separates people who do keyword research from people who fill in a spreadsheet. It takes about five minutes per keyword and it kills roughly a third of any list before a single word gets written.
- Search it, logged out, on mobile. Incognito, and set the location to where your buyers are. Your own logged-in results are personalised and lying to you.
- Count the page types in the top ten. Eight category pages and two blog posts is a ruling. Build the category page or don't enter.
- Look at what sits above the results. Ads, shopping, a map pack, an AI Overview, a video carousel. Each one takes clicks off the table before you get any, and that changes whether the keyword is worth the work.
- Read People Also Ask. It's the cheapest sub-topic research available and it tells you what the page has to answer to be considered complete.
- Check who's ranking. If the top five are all national publishers or marketplaces, and you're a six-page site, pick a different keyword this quarter and come back to this one.
- Ask what the page must contain. Not word count — components. A pricing table, a comparison, a downloadable, an author with credentials. Match the components, then beat them.
Why volume is the least useful number on the screen
Search volume feels like data because it's an integer. It isn't a measurement. Keyword tools estimate it from clickstream panels and advertising data, then round it into buckets — which is why so many keywords report suspiciously tidy figures like 1,000 or 2,400.
There are three separate ways it misleads, and they compound.
- It groups variants. Tools frequently roll close variants into one figure, so the number you see may belong to a family of queries rather than the exact phrase you're targeting.
- It measures searches, not clicks. A query answered by an AI Overview, a featured snippet or a map pack can send a fraction of its volume onward. High volume with nowhere to click is a decoy.
- It hides seasonality and it hides money. An annual average flattens a term that only exists in March. And a 90-searches-a-month query like "saas seo agency india" is worth more than a 40,000-a-month query like "digital marketing" to almost every business reading this.
Head, body and long tail — and where the money sits
Keywords sort by length and specificity into three rough bands. The bands matter because they map almost perfectly onto how winnable a term is and how close the searcher is to spending.
- The long tail isn't a consolation prize. It's usually the only band a business under ₹50 crore can win in its first year, and it converts several times better because the searcher has already narrowed the field themselves.
- Long tail also compounds. Forty pages each earning a handful of qualified visits a month is a channel; one page chasing a head term is a lottery ticket.
- A useful rule: chase body terms with your service pages, long tail with your content, and ignore head terms until your brand is the reason people search them. More on the mechanics in how to do keyword research.
| Band | Example | Character |
|---|---|---|
| Head | "seo" | Enormous volume, almost no intent, dominated by giants. Vanity target for nearly everyone. |
| Body | "seo agency india" | Real intent, real competition. Where most of a serious retainer's effort goes. |
| Long tail | "seo agency for d2c brands in bangalore" | Low volume each, high intent, individually winnable. Wins by accumulation, not by page. |
Hinglish, transliteration and the demand tools can't see
Indian search doesn't behave like the market these tools were built for. A meaningful share of real queries arrive transliterated — "seo kya hai", "gst return kaise bhare", "ghar ke liye loan" — typed in Roman script but not in English. Others come in Devanagari, Tamil or Bengali script. A growing share arrives by voice, which makes queries longer and more conversational than anything a keyword tool has in its index.
Western keyword tools underreport this demand for a structural reason, not a mysterious one: their clickstream panels skew English-speaking and desktop-heavy, and their databases are built around the phrase forms those panels produce. So the tool shows 20 searches a month for a term that's actually driving traffic to a competitor.
Two practical moves. First, trust Search Console over any third-party tool — it shows the queries you genuinely received, transliterated ones included, and it's free. Second, if you find regional demand, don't machine-translate an English page onto it. A transliterated query usually signals a different reader with different assumptions, and a literal translation reads as one. English versus regional-language SEO covers where that's worth the investment and where it isn't.