Content Research: Finding What Already Works Before You Write

Most content teams pick a topic, write 1,500 words, hit publish, and wait. Nothing ranks, because a dozen pages already cover the same ground better. The fix is not writing harder — it is doing content research first, so you only write when you have proof that demand exists and a specific gap you can fill.

What content research actually is (and what it is not)

Content research is the process of finding topics that are already proven — pages that rank, questions that get asked repeatedly, formats that consistently perform — and then deciding how you will be genuinely different from what exists. The working principle is simple: prove the need first, then do it better, instead of guessing at a topic and hoping.

It is not keyword research, though the two overlap. Keyword research tells you search volume and difficulty. Content research tells you what the winning pages actually contain, why they win, and whether the winner is beatable at all. A keyword can have 5,000 monthly searches and still be a waste of your quarter if every result on page one is a domain with hundreds of thousands of referring domains.

It is also not a content calendar exercise. Brainstorming blog ideas in a meeting produces topics nobody searched for. Content research replaces opinion with evidence.

The five sources worth your time — and what each one actually tells you

Each source answers a different question and has a different failure mode. Use at least two before committing to a topic.

1. Live search results

What it tells you: who you are actually competing with, what format wins, and how deep the current coverage goes. Search your target phrase in an incognito window and read the top 5–10 results properly. Note their word counts roughly, their structure, whether they cover the subtopics you would expect, and what they all miss.

Limitation: results are personalized and localized. Check from your target market’s location if that differs from yours, and remember that page one today is a snapshot, not a permanent state.

2. Keyword tools

What it tells you: demand size and trend direction. Volume figures vary wildly between tools — the same phrase can show 800 in one and 2,400 in another — so treat them as ranges, not facts. What matters more: is the trend flat, rising, or dying over a 12–24 month window?

Limitation: volume says nothing about intent fit. “Project management” has enormous volume and is unwinnable for a niche software vendor. Long-tail phrases with 50–200 monthly searches often convert 5–10x better per visit.

3. People Also Ask, forums, and communities

What it tells you: the actual language your audience uses and the questions search results do not answer. Reddit threads, niche forums, and comment sections reveal the second and third questions people ask after reading an article — the exact material for your angle.

Limitation: forum users are not always your buyer. A hobbyist subreddit’s priorities can mislead you if you sell to enterprises. Filter for questions that imply budget or purchasing decisions.

4. Your own Search Console and site data

What it tells you: queries where you already rank on positions 5–20. These are your cheapest wins — you have partial authority already. A page sitting at position 11 with decent impressions often needs restructuring, not a new article. Also mine your support tickets and sales calls: a question asked more than 3–4 times a month is usually a content gap.

Limitation: it only shows demand you have already partially captured. It cannot reveal topics outside your current footprint.

5. Competitor content audits

What it tells you: which of their pages drive their traffic, and — more usefully — which of their pages are stale. An article that ranks well and has not been updated in 18+ months is a candidate. Look for thin posts, outdated screenshots, and coverage holes in their topic clusters.

Limitation: tool estimates of competitor traffic are directional at best. Never build a quarterly plan on a competitor traffic number alone.

The research-to-decision table

Before writing anything, run the candidate topic through this check. If you cannot fill in every column with specifics, you are not ready to write.

Check	What you look at	Green light	Red flag
Demand proof	Keyword volume trend, forum question frequency	Stable or rising over 12+ months; question recurs across 2+ independent sources	Single spike from a news event; volume exists only in one tool
Competitive gap	Top 5 results, read in full	All results miss at least one subtopic your audience asks about	Every result covers everything and is updated within the last 3 months
Authority check	Domains ranking on page one	At least 1–2 results from sites of comparable size to yours	Page one is exclusively major publishers with 10x your backlink profile
Intent match	What the searcher actually needs at this query	You can serve the full intent with your product expertise attached	Intent is informational-only and unrelated to anything you sell
Freshness angle	Age of top results, outdated data or screenshots	Top results are 12+ months old or cite outdated figures	Results updated within 30 days by teams with dedicated editorial staff
Business fit	Connection to your product or service	Topic maps to a page that earns revenue or a defined funnel stage	Traffic would be top-of-funnel curiosity with no path to a buyer

The skyscraper trap: why “make it 20% longer” stopped working

The classic advice — find the top-ranking page, make yours longer and more thorough, then ask for links — made sense around a decade ago. In 2026 it mostly produces bloated pages nobody reads. Adding 20% more words does not add 20% more value; it usually adds padding, repeated definitions, and a table of contents people use to escape.

Here is the popular advice that is wrong: thoroughness is not the differentiator anymore, because AI-generated and AI-assisted content has made “covers everything” the baseline, not the edge. When anyone can generate a 3,000-word overview in an afternoon, length signals nothing. What Google increasingly rewards is an original angle: first-party data, a contrarian position backed by reasoning, a worked example with real numbers, or direct experience the competition cannot replicate.

Concretely, before writing, answer this question in one sentence: “What does this page contain that no competitor could produce by copying the top three results?” If the answer is “more words” or “better formatting,” you do not have an angle yet. Examples of real angles:

• A teardown of your own data: “We audited 214 client migrations — here are the three failure points.”
• A decision framework where competitors offer feature lists.
• A worked example with real costs, timelines, and what went wrong.
• A narrow version of the topic: not “email marketing” but “email marketing for regulated fintech onboarding flows.”

The practical rule: if your outline is a slightly reshuffled version of the top-ranking page’s headings, stop. Either find the angle or pick a different topic. Publishing a me-too page at scale is the most expensive way to learn this lesson.

What not to do

• Do not outsource the decision to a keyword tool’s difficulty score. A difficulty score of 30/100 means nothing if all five results above you are subdomains of companies with editorial teams of ten.
• Do not chase volume spikes. A term that jumped 400% last month because of a news event will be worthless by the time your article indexes — typically 2–6 weeks after publish for most sites.
• Do not research for two weeks and write for one day. For a single article, 2–4 hours of structured research is enough. Beyond that, you are procrastinating.
• Do not copy competitor structure. If readers see the same headings everywhere, they get no reason to stay on yours. Use competitor pages to find the gap, not the template.
• Do not skip the “why us” check. A topic can be proven, winnable, and still pointless if you have nothing credible to add. Your spent research hours are only recovered by the pages that convert.

A repeatable 3-hour research workflow

For teams producing 2–8 articles a month, this is enough structure without a research department.

1. Collect candidates (30 minutes). Pull queries from Search Console positions 5–20, recurring customer questions, and forum threads. Target 10–15 candidates.
2. Search each one properly (60 minutes). Top 5–10 results per candidate. Note format, depth, age, and the one thing they all miss.
3. Score against the gap table (20 minutes). Anything failing two or more rows gets cut.
4. Define the angle in writing (30 minutes). One sentence: the original element plus the specific audience. If a writer cannot start from this sentence, the research failed.
5. Assemble the brief (40 minutes). Include the angle, the subtopics competitors miss, questions pulled verbatim from forums, and a hard scope boundary — what the article deliberately will not cover.

How you will know the research worked

Pages built on genuine research behave differently. Within 4–8 weeks of publishing, you should see impressions climbing for the target query and for adjacent long-tail variations — a sign you covered the real question rather than just the head term. Within 3–6 months, look for engagement signals: time on page and scroll depth on the researched article versus your site average. A page that ranks but gets skimmed in 20 seconds usually means you matched the topic but not the intent — which is a research failure, not a writing failure.

Track one more number: the ratio of pages published to pages ranking in the top 10 after 90 days. Agencies and teams that research before writing routinely hit 40–60% of new pages reaching the top 10 within six months on sites with existing authority, while unresearched publishing often sits below 20%. If your ratio is low, the problem is almost never your writers. It is the topics you handed them.

Related reading

• Technical SEO — scope, method and what we measure.
• Content strategy
• Website Traffic Rankings: How Accurate Are Those Lists?
• Comparing Domains: Which Metrics Are Worth Looking At