Skip to main content
These are the questions people evaluating this category ask, including the awkward ones. Where the answer is “not yet”, it says so.

​
Trust and evidence

Yes. The visibility figure on your dashboard links to the answers it was counted from — “See the N answers behind this”. Each one shows which engine was asked, when, whether your brand appeared, and the full answer text with every mention of your brand highlighted in place.The tally on that screen is reached from the opposite direction to the headline, so if you count the answers yourself you arrive at the number on the card. You can also share one figure with someone who has no account, through an expiring link you can revoke.This covers the visibility figure. Share of voice and average position do not open onto their answers yet.
Almost certainly not, and this is the most common way these tools lose trust.Independent research found that asking the same question twice returns the same brand list less than 1% of the time. Your one attempt is one draw. rekkal’s figure is a count over dozens or hundreds of draws, printed with that count next to it.rekkal asks the same interface you did — signed out, from an address in your market — so the surface is the same one. What differs is the sample: yours is n = 1 and rekkal’s is hundreds, and it is also signed out, so it carries no memory of you and none of your history. See Methodology.
No. Two of the three engines are read in a real browser session, so there is something to photograph — rekkal does not keep one. What it keeps is the answer text exactly as it was produced, the sources shown beside it, and the character offsets of every mention in that text. A screenshot is harder to verify than the text is: you can search the text, quote it, and check it against the interface yourself.
No account is used anywhere, and none exists. Every question is asked signed out, in a fresh session with no cookies and no history carried over from the previous one. That is both a measurement decision — answers drawn through one warm session are correlated, and correlated draws make a confidence interval narrower than the data supports — and a risk one: there is no login for a provider to suspend.rekkal also does not solve, bypass or work around any bot-defence challenge. Where one stands between rekkal and an answer, that run is recorded as a failure and the week’s sample is smaller and says so.
Because they are measuring different things and rarely say which. Scraped chat UI versus API, one run per prompt versus many, mentions counted per answer versus per occurrence, different prompt sets — any one of those changes the number.What you can compare across tools is not the level but the method. Ask each vendor: how many runs sit behind this number, what is the denominator, and what happens when the difference is not statistically distinguishable. rekkal’s answers are on Methodology.

​
Method

Once per engine, per day. Your sample is prompts × engines × days, and it accumulates fast: 50 prompts on three engines is 150 answers a day, and over 1,000 in a week.That is why rekkal offers 7-day, 4-week and 12-week windows and defaults to comparing four weeks against the four before them, rather than yesterday against today.
A single day is a noisy reading, and rekkal treats it as one. Nothing about the cadence changes the discipline: every figure carries the sample it was computed from and a 95% confidence interval, and a change is only reported as a change when the two periods’ intervals do not overlap. A daily cadence gives you seven times the sample per week, which is what makes a real movement clear that bar sooner.What it is not is a licence to read one day as a trend. The default window is still four weeks against the four before it.
Because the two periods’ confidence intervals overlap, so this sample cannot tell that movement apart from noise. The card says No measurable change and tells you how much the point estimate moved.Widening the window narrows the intervals, and so does a larger prompt allowance — a higher plan, since a prompt set cannot be edited after setup yet. See Reading a change.
One market and language per workspace, derived from your domain at setup — a .fr domain settles on France and French. rekkal does not yet run the same prompt across several countries or languages in parallel, and does not claim to.
ChatGPT (OpenAI with web search), Perplexity (Sonar), and Gemini (with Google Search grounding). Every plan covers all three. See Plans and limits.
No, and the trial is not one either. Every account starts as a 7-day trial of the plan you choose: a card is required to start it, nothing is charged for 7 days, and on day 8 the card is charged that plan’s monthly price automatically. Cancelling before then is one click in Dashboard → Settings and charges nothing.The one thing that is genuinely free is the audit: type a domain, get a real reading on Gemini with every interval on screen, and no account, card or email is asked for.
No. rekkal pays for every query, which is why the plans have run allowances rather than fair-use wording.

​
Setup and scope

Setup generates a prompt set from your site and pre-selects your allowance across five topics, so you are not staring at an empty box. What rekkal does not have is a corpus of real user prompts or a planner that ranks prompts by search demand — those prompts are derived from your site, not observed from real traffic.
Not yet. Competitor discovery — automatically promoting brands that keep appearing alongside you — is not built. Nor is there a screen for adding competitors yourself yet, so a workspace’s tracked set cannot be populated in the product today.
Not in the product yet. Setup runs once per account, and the settings page shows your brand read-only. Corrections during setup are one click; after it, there is no way to change them yet.
Yes. Type one domain at /audit and you get a real reading — no account, no email, no card. It asks a small set of generated questions, ranks the brand against the rivals named alongside it, and prints the sample size next to every figure.That sample is deliberately small, so it sits below the reporting threshold and the page says so on screen rather than rounding the caveat away. It is an ordering and a bound, not a substitute for a tracked account.
It tells you where — the sources view lists domains that cited a competitor in answers about your category and never cited you. It does not yet generate recommendations. See Sources and gaps.
No. rekkal measures answers and citations. It does not read your analytics, your server logs or your CRM, and does not attribute revenue.

​
Account and billing

Prompts over your new allowance stop running. They are skipped, not deleted, and the data already collected stays.
Normally no: a call that never reached the provider is refunded to your allowance. A run that reached its engine is counted, whatever the answer said.Two exceptions, both narrow. A failed call belonging to a month that has already been billed keeps its charge in that month rather than being credited back to a month you were not charged in, and an answer the engine produced but rekkal could not store stays counted because the query was paid for.
Not yet — one account per workspace today. The data model supports seats; the product does not expose them.
rekkal is pre-launch. Billing runs against Stripe in test mode — do not enter real card details.