Frequently asked questions

This page collects the questions people email us most: what the scores mean, what a plan includes, and where a given number comes from. Answers are short, and each one points at the guide that goes deeper.

If something is broken rather than unclear, start with Troubleshooting.

What Trakkr measures

What is AI visibility?

AI visibility is how often and how prominently your brand appears when someone asks an AI assistant a buying question. It is not whether your page exists, it is whether ChatGPT, Perplexity or AI Overviews name you when a buyer asks who to use. Trakkr reports the ranked brands observed for your tracked questions, with daily reports for active brands. Core concepts explains the model behind it.

How is AI visibility different from SEO?

SEO is about ranking in a list of links that a person then chooses from. AI visibility is about being one of the few brands an assistant names inside a written answer, where there is no list to scroll and no snippet you control. The two overlap, because the pages and sources that rank well are often the ones models read, but you can rank first on Google and still go unmentioned by ChatGPT. Core Concepts sets out the difference.

Which AI models does Trakkr track?

Eight: ChatGPT, Claude, Gemini, Perplexity, DeepSeek, Grok, Meta AI, and Google AI Overviews. Paid tracking covers all eight without a separate per-model charge. Active brands receive daily reports, and per-model results show the coverage available for each question.

Where do the answers come from?

Trakkr sends each tracked prompt to the model providers' own APIs and records what comes back, so every score is built from a real answer rather than a simulation. AI Overviews is collected from live Google results. You can open any prompt and read the response a score was built from.

How is the visibility score calculated?

Trakkr scores your position in each answer: first place earns 10 points, second 9, down to 1 point for tenth, and 0 for eleventh or not mentioned. It sums those points, divides by the maximum possible across successful provider responses (including ones where you did not appear), then applies a square-root scale. Appearing first in half your prompts lands around 70. Metrics has the full formula.

What is the difference between visibility and presence?

Presence is the share of prompts where you are mentioned at all, with no weighting for position and no scaling. Visibility weights where you appear in the answer. A brand at 100% presence and 50% visibility is named everywhere but usually late in the list, which is a different problem from not being named.

Why is my score different on each model?

Each model has its own training data, cutoff date and retrieval behavior, so scoring 80 on Claude and 40 on ChatGPT is normal rather than a bug. A changed page does not guarantee a change in the next answer. Compare the report dates and evidence for the same prompt and model. The per-model split on the Dashboard is the first thing to check when a number moves.

Why do my score and my competitor rank disagree?

Visibility is an absolute number and rank is a position in a list, so one can move while the other sits still. You can score 80 and rank fourth if three competitors score higher. Filtering to a single model recomputes both the score and the rank against that model alone, which is why a filter can swap your position. When a number looks wrong, open the prompt and read the responses behind it.

Prompts, brands, and markets

How many prompts should I track?

Start with 5 to 20 questions your buyers actually ask, then grow the set once you see what moves. Growth and Scale normally include 50 active prompts per brand. Add-ons and custom accounts can differ; use the limit shown on the Prompts page. Inactive prompts do not use active slots. Prompts covers tags, audiences and the ideas row that suggests new questions.

What makes a good prompt?

A good prompt is a question a real buyer would type, specific enough to be about your category, and open enough that the model has to choose between brands. "Best running shoes for flat feet" works. "Best software" is too broad to mean anything, and a branded question changes the test by supplying a brand in advance. Branded questions can still be collected and scored, but keep them in a separate tag from your discovery baseline.

Do inactive prompts count toward my limit?

No. Only active prompts count against the limit. Deactivating a prompt frees a slot and keeps its history, so you can rotate a seasonal question in and out without losing the record.

Where can I read the raw AI responses?

Two places. Click any row on Prompts to open that prompt's own page, which shows the trend, the phrasings models used and the sources they cited. For the full unedited answer with brand names highlighted inline, open Competitors, switch to the Prompts tab, and expand a row.

Can Trakkr track prompts in languages other than English?

Yes. Trakkr carries more than sixty languages and regional variants, from Brazilian Portuguese to Flemish to Canadian French. Set it at Settings → Brands → Market and language → Shared question language. Models are prompted in that language and the answers are scored in it, with no translation step in between.

Can I track a specific country?

Yes. Every brand has a market, and paid plans can add more under Tracked markets in the same section, from a catalog of more than fifty countries. Each market runs as a separate analysis, so you can compare what AI says about you in Australia against the UK. Changes apply on the next run. To ask a single prompt from another country, give that prompt its own market in the prompt editor.

Competitors and citations

Why can't I type in a competitor?

The competitor list is discovered from real answers rather than configured, so it shows who models actually recommend next to you instead of who you consider a rival. That usually surfaces names you had not counted. If a brand you expect is missing, add prompts where you would expect it to come up, and it appears once a model names it.

Can I hide or merge competitors?

Yes. The row menu on Competitors has Hide competitor for names that are not really rivals, and you can unhide them later from the same place. The Groups button in the toolbar merges variants of one brand, so "On" and "On Running" count once. Trakkr suggests groups it detects, and you accept or dismiss each one. Competitors covers the rest.

What counts as a citation?

A citation is a URL a model returned as a source for an answer. Some models publish their sources outright, so those are recorded as given. The Citations page collects them under Sources, with Queries for the searches behind them and Videos for the video sources models cite. Citations explains how to read the gaps between what cites your competitors and what cites you.

How do I get cited more often?

Get onto the pages models already read for your category: round-ups, review sites, industry publications and the community threads that keep returning as sources. The Outreach tab does the finding for you. It lists publishers that cite your competitors but never you, ranked by how much visibility the gap costs rather than by how well they fit, with a researched pitch for each. See Outreach.

Acting on what you find

How do I actually improve my visibility?

Start on Pages, which walks each URL through Available, Reached, Understood, Relevant, Selected and Visited, then names the first stage holding it back. Turn that fix into an Action. When the work is done, Results measures it against a window fixed at the start and tells you what it earned.

How long do changes take to show up?

Perplexity and AI Overviews read the live web, so a fixed page can change an answer within days. ChatGPT, Claude and the rest lean on training data and usually take weeks to months. In Futuro's record, ChatGPT first named the brand for its category term eight days in, and one fixed question took twelve weeks to hold.

What is a page record?

A page record is everything Trakkr knows about one URL: the six journey stages, the current bottleneck, suggested actions, work in progress, measurements, history and results. Open one when a site-wide number tells you something is wrong but not which page to fix.

What does "Couldn't measure" mean?

It means the measurement window closed without enough page-level evidence to judge the work. The usual causes are no target page, no before reading, no connected source that can see the page, or no activity on the page in the window. It is not a failure and not an error, and it is excluded from earn rate rather than counted as a loss.

Why does Trakkr say "coincided with" instead of "caused"?

A before-and-after window can show that a metric moved after work shipped. It cannot rule out the launch, the season, or the competitor who shipped the same week. Trakkr reports the timing and the size of the change and leaves the causal claim alone.

What is the difference between an action being done and having earned something?

Done means the work happened. Earned is a later verdict: a page-level metric cleared the bar that was fixed when the action started. In between, the result reads Measuring. The two never merge, so a busy month of shipping cannot pass itself off as a good month of results.

What is AI Pages?

AI Pages serves a second, machine-readable version of each page to AI crawlers: the same facts, restructured with schema in the head, a summary and key facts in the body, and tracking scripts stripped out. Humans and search crawlers get your normal site. It runs as a small piece of middleware on your own hosting, so nothing about the site changes for anyone else. See AI Pages.

What does Perception tell me that visibility does not?

Visibility counts whether models name you. Perception reads what they say about you: the words models reach for, the scored dimensions behind them, and the quotes each score came from, set against competitors you pick. Use it when you are named often enough but described wrongly, or when a positioning change needs to show up in the answers.

What are Automations?

Automations are work that runs without you. That covers exact "when this happens, do that" rules, such as posting to Slack when visibility drops, and Agent judgment calls that prepare work inside the permission you granted. Every automation still answers to the same six permission areas, so nothing ships that you did not allow.

What can the Agent do without asking?

Nothing that changes anything. The Agent reads your data and answers questions by default. Preparing or making a change needs a permission you grant per work area, and the six areas (Technical fixes, Page improvements, Content refreshes, New content, Getting listed, Community replies) are all off until you turn them on.

Plans and billing

What plans are there?

Growth is $100 a month and covers 1 brand, 50 active prompts, all eight models, a team of three including you, and MCP access. Scale is $500 a month and covers 10 brands, 50 active prompts per brand, unlimited team seats, the REST API, and the white-label client portal. Annual billing is ten times the monthly price. Enterprise is custom. Plans and billing has the rest.

Is there a free version of Trakkr?

No. Trakkr runs a 14-day Growth trial that needs a card and charges nothing until the trial ends, and after that every plan is paid. The free tools on the marketing site sit outside the product and need no account.

Can my whole team use one account?

Share the brand instead of the login. Growth includes a team of three, counting the owner, and Scale has unlimited seats. Extra Growth seats are $20 a month each. Members inherit the owner's plan, and only the owner (or an admin the owner grants billing access) can change the subscription. Teams covers roles and invites.

Can I add more brands or prompts?

Yes, both are add-ons under Settings → Billing. An extra brand is $50 a month on current subscriptions, and older subscriptions may still pay $39. Extra prompts add 50, 100 or 200 active prompts per brand, priced per brand. Billing previews the exact change and the new total before you confirm it.

How do I cancel?

Go to Settings → Billing and use the Cancel subscription link at the bottom of the page. Access runs to the end of the period you have already paid for, then paid features pause and your data stays put so you can pick it back up. If you are on a team you do not own, that same row names the owner to ask instead.

What if I need more than Scale offers?

Add-ons cover most of it. Extra brands, extra prompts, extra markets and extra article credits all attach to a Scale subscription without a call. Past that, Enterprise covers custom limits, SSO, invoice billing and custom terms. Email [email protected] with your team size, brand count and integration needs.

Can an agency run client brands in Trakkr?

Yes. Agency mode turns a Scale account into a portfolio: one roster of client brands, cross-brand work, scheduled client reports, and a white-label client portal on your own domain and branding. The portal is included with Scale rather than sold as an add-on, and portal users see their own brand only, read-only. See Agency.

Does Trakkr have an API?

Two of them. The MCP server comes with any paid plan and lets ChatGPT, Claude or Cursor query your data conversationally. The REST API is included with Scale. Both are set up at Settings → Developer, and the reference lives at /learn/api.

Data, access, and help

Is my data private?

Your prompts, brands and results are private to your account. Trakkr never sells your data, and your brand data is never used to train AI models. The privacy policy and security page have the specifics, and security questions go to [email protected].

What does my developer need to connect my site?

Usually a URL and one OAuth click. WordPress connects over the standard /wp-json REST API using OAuth or an application password, Vercel and GitHub connect through their own apps, and crawler tracking has a copy-paste snippet for more than fifteen platforms. No IP allowlist is needed, because every connection is ordinary HTTPS from managed infrastructure. See Install crawler tracking.

How do I change the email on my account?

Open Settings → Profile and use Change email address. Trakkr sends a confirmation to both the old and the new address, and the change completes when you click the link in either one. Brands, history, team membership and billing all carry over. If you signed up with Google, keep using Google, and email [email protected] if you need to move to a different identity.

Can I get my data out of Trakkr?

Yes, four ways. Reports keeps the run history per brand, CSV export pulls a scoped slice, Google Sheets and Looker Studio keep a live connection, and the REST API on Scale gives you the raw records. Reports and exports shows which one suits which job.

How do I get help?

Click Support at the foot of the sidebar. It offers Read the docs, Ask the Agent, and Message support, which is live chat when someone is around and email when they are not. You can also email [email protected], or book a demo if you would rather be walked through it.