Hook: The model writes the buyer questions your team never wrote down, asks them, and names every brand that answered them. Question research for one market costs EUR 720 to EUR 2,160 of specialist time.
What You'll Be Able To Do
- Turn a campaign brief into the buyer questions nobody in your building wrote down
- Size a run with three numbers and know what each one buys you
- Read the Brands Detected ranking, with a count, an average, a range and a variance for every spelling of your own name
- Turn a topic gap into a content plan with briefs attached, without leaving the tab
- Take three exports away: the CSV brand table, the Markdown with the topic descriptions the screen withholds, and the PDF you send
Runtime: 6:43 | Difficulty: Advanced
You own the prompt set your brand is measured against. You wrote it, or people who already know the product wrote it, so every question on it is a question your team knew to ask. The blind spots are the ones nobody wrote down, and they stay invisible until a customer asks one and a competitor answers it. A keyword tool will not find them either: it reports what people typed into a search box, and says nothing about the questions a model writes when it reasons about your category on its own.
What it costs today
Assumption: an SEO specialist at EUR 90 an hour.
| Job, by hand | Time | Cost |
|---|---|---|
| Question and keyword research, one market | 1 to 3 days, so 8 to 24 h | EUR 720 to EUR 2,160 at EUR 90/h |
| Research seat (Ahrefs, Semrush, AnswerThePublic) | ongoing | EUR 99 to EUR 199 a month |
That price is per market and per language. Language is a field on the form, so a second language is another run on the same subscription.
Discovery Lab sits on Strategist at EUR 496 a month, two and a half to five times a research seat (496 / 199 = 2.5, 496 / 99 = 5.0). What the EUR 496 buys is the part the keyword seats have no way to produce: the questions a model writes on its own, the brands it reaches for when it answers them, and a content plan built off the gap you pick.
What the video shows
- Discovery Lab opens on the run log in the history dropdown: persona, industry, market, date, and the three numbers that produced each run. Read it before you build anything, because the fastest way to design a run is to copy the last one and change one thing.
- The campaign brief goes into the context box and the counter passes the 50-word minimum. Name the product, the market, the price band and who you are up against, because the model writes the questions from this paragraph.
- Four fields frame the market: language, industry, market context, persona. Each one changes a different part of the output, and language alone makes the same brief two different reports.
- Every spelling of your brand goes in as its own chip, then the seed keywords. Anything you leave out gets counted as somebody else's brand.
- Three sliders size the run. Answers = prompt count x (1 + mutations) x iterations, so the 5 / 2 / 3 defaults give 45 answers and the 20 / 5 / 10 ceiling gives 1,200.
- Press Run and walk away. The button reads Queued, then Running, and the job survives a page reload.
- The Insights tab opens on four counts, then the topic groups: the messages that came back often at the top, the gaps that appeared once at the bottom. Claims are compared by meaning, so the same point worded two ways lands in one group.
- Brands Detected gives one badge per spelling you typed. Click a badge and four cards swap: total mentions, average per answer, the range between the quietest and loudest answer, and the variance. Every brand the model named sits underneath.
- Hover a gap row, click Generate Content, and the topic turns into a plan with up to six briefs attached. Do it on a second gap and the second plan lands under its own heading.
- Reload an earlier run from history, change one number, run it again. Two runs that differ by one variable tell you what the setting was worth.
What you need
- Strategist, EUR 496 a month, or Enterprise.
- A project with a completed scan and at least one persona.
- A campaign brief of 50 to 500 words. The run button stays off below 50 words.
- At least one keyword. Brand spellings are optional and worth typing.
- A persona picked on every run. The label reads optional and the server requires it.
- One discovery run and one content plan at a time per scan.
Questions people ask
Does this replace Ahrefs or Semrush? No. Keep the keyword seat for search demand; this run covers the questions a model writes on its own and the brands it names against them. Plenty of teams run both.
Can one run cover two personas? No. The form takes one persona per run, so two personas means two runs. The same goes for two languages.
How big can one run get? Twenty prompts, five rewrites each, asked ten times: 1,200 answers, and a long wait. The defaults are 5, 2 and 3, which is 45. Iterations is the number that decides the shape of the report, because at one iteration nothing can repeat and the split tells you nothing.
What does the frequency percentage next to a brand mean? Distance from the top brand, measured as its share of the leader's mentions. Read it as share of voice and you will misreport the run.
Bring the brief you would hand an agency, and let the model write the questions your team never asked.