Turning research into landing page copy that sounds like the customer

The short answer: build the page from four lines you did not invent — the problem in the audience's own words, the workaround they currently use, the point where that workaround breaks, and what becomes possible after. All four should be harvestable from your source material as near-quotes. If you cannot point at where a line came from, it is very likely a line you wrote for yourself.

The usual failure is not bad writing. It is that the page starts one step past where the reader is: it opens on what the product does, to someone still deciding whether this is even about them.

Harvest phrasing before you write anything

The first pass produces no copy at all. It produces a list of sentences people actually said, collected from material where they were explaining their own situation rather than selling to anyone.

Demonstrations and walkthroughs are unusually good for this. Somebody narrating a process to a general audience uses the vocabulary that is natural to them, at the moment of the friction, unedited. That is the register your headline needs, and it is almost never the register that emerges from a whiteboard session with the founding team.

The right-hand column is not more clever. It is more specific, and specificity is what makes a stranger conclude that you have met someone like them. Collecting it is a mechanical job — the same corpus you built to decide what to build already contains it, which is why copy research rarely justifies a separate project.

The four lines that carry the page

Nearly every page that converts contains these four moves in order, and each maps to something in the research.

Features come after all four, and they belong there — as evidence for a claim the reader has already accepted, rather than as the claim itself.

Respect the workaround

The strongest temptation in research-driven copy is to attack the thing people currently use. It is nearly always a mistake. The workaround is working, at their current scale, and telling someone their functioning system is bad reads as not understanding it.

The move that works is agreement followed by a boundary: the workaround is genuinely fine, here is the specific point at which it stops being fine, and that point is a place your sources actually described. A contented workaround is also the signal that most often ends a product idea before the page is ever written, which is worth checking first — see how to kill a SaaS idea early.

Specificity is the whole mechanism

The reason harvested lines outperform written ones is not authenticity in any romantic sense. It is that real complaints carry numbers, tools and durations, and invented ones carry adjectives.

“Save time on research” asks the reader to supply the content. “Nine hours of videos and still no plan” supplies it, and the reader either recognises the situation or does not — which is the decision you actually want them to make at the top of the page. A headline that nobody can disagree with is a headline nobody can identify with either.

This has a practical consequence for what you harvest. Prefer the source sentence that contains a quantity over the one that expresses the same idea generally, even when the general one is better written. The number is doing the work, and it is also the part you could not have invented credibly, which is why sourced numbers must stay attached to where they came from rather than being rounded into something more convenient.

Where category language belongs

The landing page headline is the wrong place for a category label unless the audience already shops in that category. “Research platform” means nothing to someone who has never bought one, and spending the headline on it buys confusion at the most expensive position on the page.

The category earns its place further down and especially on the pricing page, where the visitor has switched from “is this about me” to “what am I comparing this against”. That is also where research feeds the numbers rather than the words, using what the material says about existing spending — the approach in pricing research from public content.

Objections belong on the page, in their order

Research produces one thing most copy processes never get: the objections in the order people actually raise them. Sources hesitating about a category say the same three or four things, and they say them in a reliable sequence — feasibility first, then effort, then trust, then price.

A page that answers them out of order loses people early. Leading with price to someone who has not yet accepted that the thing is possible is answering a question they have not reached; putting the trust material at the very bottom means the people who needed it left three screens ago. The sequence is not a matter of taste, and the corpus hands it to you for free.

The pattern to look for is what a source says immediately before they stop. In a walkthrough, the sentence just before someone abandons a tool is usually the objection in its purest form — unhedged, specific, and said to nobody in particular. Those moments are worth timestamping deliberately, because they are the shortest route from research to a block of copy, and they are also where the recurring frustrations that make a whole product surface, as described in extracting product ideas from video. Collecting those moments systematically, rather than as they happen to turn up, is its own short research pass — finding the objections your buyers already voice.

Testing whether the copy is really theirs

Two checks catch most self-written copy before it ships.

The first is attribution: for every line above the fold, point at the source it came from. Lines with no source are not necessarily wrong, but they are the ones to rewrite first — they are where the founder's vocabulary leaks in.

The second is the swap test. Replace your product name with a competitor's. If the page still reads as true, the copy describes a category rather than a product and will convert like a category page — which is to say, poorly. Positioning that survives the swap test usually comes from a teardown of what the alternatives actually fail at, covered in running a competitor teardown from public video.

What the research half costs

Copy research is lighter than build research because you are collecting vocabulary, not proving demand. Eight to twelve sources is usually enough, and the bar is repetition across independent voices rather than volume.

As of August 2026 that fits the $19 Hobby plan — 2 active projects, 25 videos a month — with room for a second pass when the page underperforms. Agencies writing pages for several clients are the case for Studio at $199 with 20 projects and 250 videos, which is the same shape as running client market research at agency scale. Tier detail sits on the pricing page.

Closing thought

Copy that sounds like the customer is not a writing achievement, it is a listening one. The sentences are already out there, said plainly by people who had no idea they were writing your headline — and the only real skill involved is resisting the urge to improve them.