Research

Most people hear "AI brand research" and assume we mean SEO research with a new coat of paint. We don't.

Research in Trakkr is sampling. We take the kinds of questions your buyers ask AI, ask those questions to the real models on your behalf, and capture exactly what comes back: which brands get recommended, in what order, with which sources cited. The output isn't a keyword report. It's a snapshot of how AI talks about your category right now, with you in it (or not).

That distinction matters because the old playbook breaks down here. Typing a few queries into ChatGPT yourself isn't research, it's anecdote. The model gives different answers on different days, in different sessions, with different phrasings. You need volume to see the pattern, and you need it across the models your buyers actually use, not just the one you happened to open. That's what a Research run does in about five minutes.

Research is not SEO research

The unit is different. The mechanic is different. The decision it informs is different.

If you walk into Research expecting a keyword tool, you'll under-use it. Walk in expecting a market-position scanner and the page makes immediate sense.

The two views

The Research page has two tabs, and they answer different questions.

• Full Research is the broad scan: roughly 500 generated questions across your whole category, so you can see where you stand overall and find prompts worth tracking.
• Topics zooms in: every theme in your market with your estimated standing in each, plus the option to deep-scan any one of them for a sharper, fresher read.

Most people start with a full scan to map the territory, then use Topics to go deep on the slices that matter. We'll walk through both.

Running a full scan

You hit Run Research. About five minutes later, you get a complete picture of your category.

Behind the scenes, Trakkr generates roughly 500 {discovery questions|Open category questions where AI has to choose which products to recommend, the kind buyers ask before they've settled on a brand.} tuned to your brand, industry, competitors, and description. It asks those questions across the AI models, parses every response, and stitches the results into a single view.

A few things worth knowing about how the sampling works:

• The questions are generated fresh, not pulled from your tracked list. Trakkr writes them from your brand description, industry, and known competitors. They're a separate, independent set, not your active Prompts. The more complete your brand profile, the sharper the questions. Thin profile, generic questions.
• Each question is sampled once. A full scan takes a single pass per question to map the field quickly. Your Prompts page is the opposite: it runs every tracked prompt against all eight models, every day. Same question, different lens, which is why a Research position and a Prompts score for the "same" prompt won't always line up.
• It's a snapshot, not a tracker. A full scan captures one moment, and re-running replaces the last one. Anything you want to watch over time belongs on your Prompts list. You can refresh a full scan about once a week.

Reading your results

When the run finishes, the page stacks a few things together. Read them top to bottom and a story tends to emerge.

The headline stats

A row of four numbers sets the scene: Analyzed (how many questions the scan covered), Appearing In (how many you showed up in, with the percentage), Top 3 Positions (where you're a first-pick), and Opportunities (the questions you don't appear in yet).

Your position distribution

A donut chart (Your Position Distribution) breaks every question in the run into three buckets: Top 3, #4-10, and Missing. The shape tells you what kind of market you're in.

Click a slice to filter the table underneath, so you can read the questions behind that bucket.

Market leaders

Next to the donut, a ranked list of the brands AI mentioned most often in your run, each with a {share of voice|How often a brand shows up in your scan, measured as a share of all brand mentions.} bar and your own count alongside. If a competitor you've never heard of shows up at #2, that's a thing to know.

The insights strip

A thin strip above the table highlights statistically interesting patterns from this run. Trakkr looks for slices where your performance diverges meaningfully from the baseline and surfaces them as chips.

Click any chip to filter the table to the questions behind it. That's usually faster than scrolling.

The question-by-question table

Every row is one question, with your rank position, the estimated {AI search volume|An estimate of how often a question gets asked across AI tools, measured where we have data and modeled where we don't.} for it (the AI Vol dots), and, on expand, the competitors that appeared in the response. The filter tabs at the top mirror the donut slices, plus an All view and a Close view (your #4-10 questions, sorted by how near you are to breaking into the top 3). Sort by Rank, AI Vol, or Prompt, and use the search box to find a specific question.

Position colour reads at a glance:

From research to tracking

The whole point of a full scan is to decide what to start monitoring. Two paths.

Quick Wins. When a run completes, the Quick Wins button in the header collects the questions just outside the top 3 (your #4-5 and #6-7 positions) where the smallest amount of work moves you furthest. These are the high-volume, close-but-not-winning questions. Review them in a modal, select the ones you want, and add them to your active list in a click. If you only do one thing after a run, do this.

Pick by hand. Browse the table and click Track on any row to add that question to active tracking. Use the checkboxes to add several at once. Filter to Missing to focus on coverage gaps, or Close to hunt for movable rankings. Where you're not appearing, Diagnose opens that question in the Diagnose tool to explain why.

Whatever you add starts running through Trakkr's daily pipeline the next morning. The scan itself is one-and-done (you can also Export it to CSV), but the prompts you promote out of it run forever.

Topics: zooming in

Sometimes you don't want your whole market. You want to look hard at one slice of it: enterprise pricing, open-source alternatives, a vertical you're moving into, a specific competitor's territory. That's the Topics tab.

Topics shows every theme in your market and how you show up across each one. The key idea is that every topic comes in one of two depths:

• Estimated topics are rolled up for free from your last full scan. They're instant and cost nothing, a quick read on where each theme sits using questions you've already run.
• Deep scans go further. A deep scan researches 50 fresh AI questions about one topic and measures exactly how you show up there, in about 3-5 minutes. It's narrower and sharper than the full scan, and it costs one deep-scan credit.

So the tab gives you a free, estimated map of every topic, and lets you spend a credit to get a precise, current reading on any one that matters.

The topic list

Each row shows the topic, your mention rate (how often you appear in it), your avg pos (average ranking), and a visibility bar. A small badge marks whether a row is estimated or a real deep scan. Sort by any column to find your strongest and weakest themes.

To go deeper, click Add topic: give it a name (a sub-category, use case, or phrase narrower than your market) and an optional line of context to steer it, then Run deep scan. Trakkr suggests topics drawn from your own research, and a Topics from your research you haven't explored rail surfaces themes you've seen but never zoomed into.

Inside a topic

Click any topic to open its drawer, which reads like a focused mini-scan:

• Run history — deep scans accumulate. Each time you re-run a topic, it joins a timeline so you can watch a theme move over weeks. Click any point to view that run.
• Sub-topics — Trakkr auto-tags the slices inside a topic (for example "fee-only advisors" inside "financial advisors") so you can see which corner of a theme you win or lose.
• Best and missed questions — the questions where you rank highest, and the ones where competitors appear and you don't.
• Market leaders and position distribution — the same donut and leaderboard as the full scan, scoped to this topic.

Deep scans persist; full scans replace

This is the practical difference. A full scan replaces the previous one, so you only ever have the latest. Deep scans are saved and build a per-topic history. If there's a slice you want a durable, point-in-time record of, deep-scan it rather than relying on the full run.

Your monthly deep-scan allowance is per active brand, so the more brands you actively track, the more deep scans you have. The quota resets on the 1st of the month.

Research vs the Prompts page

These get confused a lot. The shortest version: Research is discovery, Prompts is monitoring.

You don't pick between them, you use them in sequence. Run Research to widen the field. Promote a handful of questions to your Prompts list. Watch those daily. Re-run Research when the market shifts.

When to run it (and how often)

Research is a periodic activity, not a continuous one. Run it when something material has changed:

• Setting up a new brand. Always start here. You don't know what to track until you've seen the market.
• After adding or losing a major competitor. Their presence changes which questions are interesting.
• After a launch, rebrand, or pricing change. Verify AI has caught up to your new positioning.
• Quarterly, even if nothing changed. AI training data shifts, models update, your category moves. A quarterly scan catches drift.

Re-running a full scan is non-destructive in the sense that you never lose your tracked Prompts, but you do lose the previous full snapshot. If you want to preserve a point-in-time view of a slice, deep-scan it in Topics, those stick around.

Common questions

How does Trakkr decide which ~500 questions to generate?

It reads your brand description, your declared industry and competitors, and your website, then asks an AI model to generate discovery-style questions across a balanced mix of intents (recommendation, comparison, best-for, and so on) and focus areas (general, budget, enterprise, features, integrations). Fill out your brand profile properly and this gets sharper.

Are these the same as the prompts I'm tracking?

No. A full scan generates its own independent set of roughly 500 questions for the run. It doesn't read from, change, or consume your active Prompts list, the two are entirely separate. Research is where you find questions; your Prompts list is the handful you've chosen to track. Questions only cross over when you click Track to promote one.

Why is my Research position different from my Prompts score?

Because they measure different things. A Research run samples each question once. Your Prompts page runs each prompt across all eight models and averages, every day. Same question, different lens. If you add a question from Research to your active list, expect the score to shift once it starts running across the full model fleet.

What's the difference between an estimated topic and a deep scan?

An estimated topic is rolled up for free from questions in your last full scan, a quick read at no cost. A deep scan runs 50 fresh questions focused on that topic for a precise, current measurement, and costs one deep-scan credit. Estimated tells you roughly where a theme sits; a deep scan tells you exactly.

Does running Research affect my tracked prompts or quota?

No. Research is a separate run with its own machinery. Your active Prompts keep running their daily cycle untouched, and Research doesn't consume prompt slots. Deep scans draw from their own monthly allowance, not your prompt limit.

Can I re-run Research without losing my last one?

The full scan replaces the previous one each time. If you want to preserve a specific view, deep-scan that slice in Topics instead, deep scans are saved and build a history per topic.

Why did Quick Wins suggest a question I'd never track?

Quick Wins ranks by demand and proximity, not strategic fit. A high-volume question where you rank #4 will surface even if it's tangential to what you sell. Treat the list as a short list, not a checklist.

Topics or deep scans aren't available on my plan, what's the gate?

The whole Research feature is paid. Growth and Scale both include 5 deep scans a month on each active brand, and Scale trials get 1 to try. Deep scans only run on an active (tracked) brand. Free accounts see a preview. The allowance resets on the 1st of the month.