Hook: Someone in the meeting asks which AI names your brand, and the only evidence in the room is a screenshot from one chat window. Tracing that single question properly by hand costs EUR 180 to EUR 360 of analyst time.
What You'll Be Able To Do
- Put one buyer question to Gemini, ChatGPT and Grok in a single run
- Read the searches each model fired before it wrote its answer, and the pages it read
- See which brands survive when the model rechecks its own work, and which drop out
- Compare all three models in one picture, brand by brand
- Turn the gaps into content plans built from the model's own searches
- Hand a client, a CEO or a prospect a link they open with no account
Total Time: 6:29 | Difficulty: Advanced
Your buyer never stops at one answer. They read it, then ask about the brands that answer named, and every hop drops somebody. Doing the job properly means running the question on three models, following every brand each answer names, and logging which of them survive the next hop. Almost nobody does it, because by hand it takes hours per question.
What it costs today
An analyst at EUR 90 an hour. Tracing one question by hand means running it on three models, following every brand each answer names, and logging which of those brands survive the next hop.
| Line | Rate | Hours | Cost |
|---|---|---|---|
| One question, three models, by hand | EUR 90/h | 2 to 4 | EUR 180 to EUR 360 |
| A 20-question set | EUR 90/h | 40 to 80 | EUR 3,600 to EUR 7,200 |
Forty to eighty hours is one to two weeks of one analyst. Half of that work cannot be bought at any price: no consumer chat window shows the searches a model fires before it writes its answer, so nobody reconstructs them by hand.
Answer Trail adds nothing to the bill. It carries no tier gate, and the whole three-model group counts as one job, so three models spend the queue slot of one.
What the video shows
Eight shots, in order, so a reader who cannot watch still learns the flow.
- On the Answer Trail card, switch to Compare models, type the buyer question, name your brand, set the region, and leave the persona on General audience.
- The three runs go out together and land apart. Each card names how that model captured its answer: Gemini through grounding, ChatGPT through web search, Grok through its API.
- Open the Gemini trail. Every search the model fired before it wrote a word is listed, badged Intent or Site probe, over the pages it read, grouped by domain.
- Brand verification. The model writes a fresh search per brand and grades it Survived, Dropped, Emerged or Absent. Dropped is a proof problem. Absent is a coverage problem.
- The same question, one model over. A different fan-out, different sources, and a different set of brands surviving.
- Cross-model folds the three runs into one picture: your average presence at the top, then one row per brand cut into one segment per model.
- The Prompts tab turns the model's own searches into a brief. Untick down to the gaps, press Produce, and each one becomes its own content plan.
- Share mints the link, public or locked with a PIN, with an optional campaign tag. Every run stays in the history dropdown, so you reopen it mid-meeting.
Behind each verdict sits the Signals bar, which splits the citations four ways: other independent, review, tier-one news and brand-owned. A bar that is mostly brand-owned means the model is quoting you about you.
The three models disagree, and the disagreement is the useful part. A brand can hold Survived on Gemini and sit at Absent on Grok. Green on one model and red on two points at coverage. Amber on all three points at content. The two findings buy different work, and one screenshot from one chat window hides the difference.
What you need
Any plan, including Free. Answer Trail carries no tier gate.
Before you run:
- A project with a finished scan, so the persona list has rows in it. Free and Marketer allow one project per workspace.
- Member access to that project. A share link minted from outside it is refused.
- A commercial question, written in the language of the market you are asking about. The prompts that come back follow the language of the question, never the language of the Region field.
- The persona decided before you press run. Every content plan you produce later inherits the run's persona and model, and nothing re-points a plan afterwards.
Questions people ask
Which models does it use? Gemini, ChatGPT and Grok. Three, and only three. There is no Claude and no Perplexity anywhere in the run.
Does picking a persona change what the models answer? No. The trail runs the same either way. The persona rides into any content plan you produce from the run, so choose it before you press run.
How many runs can I have going at once? One per project, and a three-model comparison counts as one. A second run started before the first finishes is refused, so a twenty-question set goes through one at a time.
Does Rankfor.AI watch the question and tell me when the answer changes? No. There is no watcher, no schedule and no alert. A trail is a run you start yourself, and the link is a snapshot of that run, dated on the page. To see movement, run the same question again and compare the two links.
Does a produced plan give me a finished page? No. An Answer Trail plan is an outline plus SEO. The assets get generated after that.
One run answers the question your room is arguing about, and the link goes out the same afternoon. Start with Rankfor.AI.