Keyword research without a search volume: what live results can tell you

By Thomas Lovas Lokoy · Published 16 October 2026 · 5 min read

Every keyword tool shows a number next to every phrase, and that number has trained a generation of writers to believe keyword research means ranking phrases by volume. It does not have to. This guide explains what the volume figure is, why you can do useful research without it, and what you can read straight off a results page that a volume number will never tell you.

What a search volume actually is

No public tool has Google's query logs. The monthly figure in a keyword tool is an estimate, produced from a mix of advertising-platform ranges, clickstream panels and the tool's own model, and different tools give different figures for the same phrase on the same day. That does not make the numbers useless. It makes them estimates, and an estimate presented as a measurement is the kind of fact that quietly misleads a content plan.

AI Orbit takes a deliberately strict line here. Its research step is instructed never to state a search volume, a difficulty score or a cost per click, because it cannot measure them, and the brief panel on every article says so in one sentence. The reasoning is simple: a made-up number is worse than no number, and a tool that produced one would be inventing a fact in the one place where facts matter most. What it does instead is the subject of the rest of this guide.

What a results page tells you directly

Open a search for your topic and look, without a plugin, at what is actually there. Six things are readable by anyone:

1. The titles that rank. They show you how the question is phrased by the people who have already won it. If every title says "payback period" and yours says "return on investment", you are using the wrong words.
2. The subheadings on those pages. They are the sub-questions the topic is understood to contain. A page that ranks for a broad phrase and has fifteen H2s has done the decomposition for you.
3. The question forms. Search the "how", "why", "is it worth it", "vs" and "best for" versions of your phrase. The results for those are different pages, and the differences show you which intent each phrase carries.
4. The "people also ask" questions. These are questions, in searchers' own words. They are the cheapest list of H2 candidates in existence.
5. What the top results do not cover. Read three of them and write down the question you still have. That gap is the only reason a new article deserves to exist.
6. The language of the market. The phrase people use is often not the phrase the industry uses. A results page shows you the searchers' version.

None of this needs a volume. All of it is evidence you observed rather than a figure you were told.

How AI Orbit does it

On Pro and above, the research step is a separate model call with web search enabled and a cap of six searches. It is instructed to run exactly the process above: search the topic, look at the titles and subheadings that rank, search the question forms, and note what the top results fail to cover. Every phrase it reports must be one it saw in a real title, heading, "people also ask" block or body text. It returns a brief with five parts:

• the single primary phrase for the article to target,
• four to eight supporting phrases for headings and body copy,
• three to six questions searchers ask, phrased as they ask them,
• one to four gaps the current top results fail to answer,
• one sentence on the angle that makes a new article worth publishing given what already ranks.

That brief is then folded into the writer's instructions: the primary phrase goes in the H1 and the first hundred words, each question becomes its own H2 where it fits, the gaps are to be filled, and anything that reads like SEO is to be rewritten. You can run the research on its own with "Preview the keyword brief" before spending a credit, and the finished article shows the brief it was written from. The pipeline tutorial covers where it sits in the run.

Doing it by hand in twenty minutes

If you are not on a plan with research, or you want to check what a tool gives you, the manual version is short:

1. Search your topic sentence. Copy the ten titles.
2. Search the "how", "why", "vs" and "best for" variants. Note which ones return different pages; those are separate intents.
3. Copy the "people also ask" questions from each.
4. Open the top three results and skim their H2s. Write the one question you still have.
5. From all of that, choose one primary phrase (the one the ranking titles use), five supporting phrases (from their H2s), and the questions you will answer. Write the gap as a sentence.

You now have a brief that cites what you saw, and you can defend every line of it to an editor by showing them the page.

What you lose without a volume, honestly

You lose a way to compare two topics by size. If you must choose between writing about A or B and have no other signal, an estimate of relative size is some help, and the manual method above does not give you one. What it does give you is a reading of intent and of the gap, which decides whether the article can win at all. A phrase with a huge estimated volume and ten complete answers already ranking is a worse bet than a smaller one nobody has answered properly. Size tells you the prize; the results page tells you whether you can reach it.

Turning the brief into the article

Three rules make the brief useful rather than decorative. Use the primary phrase where a reader would expect the subject to be named, the title and the opening, and nowhere mechanical. Answer the questions outright, as their own sections, in the words the searchers used. And build the article around the gap, because the gap is the only part of the brief that explains why anyone should read this page rather than the three that already exist.

If you want to see what a brief looks like for your own topic before committing to anything, the research preview is on Pro and above; the pricing page has the plans, and how it works is the two-minute version.