Privacy policy

Who we are and what this covers

Vercatus (Oktelo Solutions), Budapest, Hungary, provides the website at vercatus.com, the workspace at app.vercatus.com and the API. This policy covers all three, and the emails we send. It is written for the people who visit the site, the people who use a workspace, the people whose details appear on the websites our customers analyse, and the people named in the AI answers we record or on the public pages those answers cite.

Controller and processor

For the data about you as a visitor, a user or a billing contact, we are the controller: we decide why and how it is processed, and this policy is the notice the law requires.

For the content of a website a customer asks us to analyse, and for the answers the engines give to that customer's questions, we are the processor: the customer decides what to measure, and we process that data only on its instruction. The data processing agreement governs that relationship and applies to every workspace.

For the public pages that answer engines cite, which we keep once for every customer with nothing about who asked, and for the figures we pool across customers, we are the controller, on the basis of our legitimate interest (Article 6(1)(f)) in knowing which public pages answer engines rely on and in improving the service for every customer; the sections below describe them, and you can object to either ("Your rights").

What we collect, why, and for how long

Each row is one category. The legal basis refers to Article 6 of the GDPR; the retention column is the rule, and the retention statement below is its summary.

We do not collect special categories of data, and we do not make automated decisions with legal effect about anyone. We do not sell personal data and we do not share it for advertising. Visitors to vercatus.com are not tracked across sites or identified: the site sets no cookie, and counts page views with Cloudflare Web Analytics, which uses no cookie and no device fingerprint.

Content of the sites we analyse

A run fetches publicly available pages of a domain the customer has verified by DNS record or meta tag. Those pages may contain personal data the site publishes, such as an author's name, a contact address or a quoted person. We process it only as part of the customer's analysis: the text is read by the language models named below to find what engines can read and to write the plays, and passages may be quoted in the customer's workspace as evidence. We do not extract people from pages into a directory, do not enrich them, and do not use page content to train models.

If you appear on a customer's website and want to know what was processed, ask the site's owner, who is the controller, or write to us and we will pass the request on.

Answers, briefs and the pages engines cite

The poll. For each site on a plan, Vercatus asks the answer engines listed below the questions a market asks on the site's topics, every day, with each engine's web search on and, for some engines, off. The questions are written from the site's profile, from the workspace's tracked prompts, from the searches engines ran while answering and, when the site's Search Console is connected, from its buyer, problem and brand searches (see "Data from your Google account"); the workspace can add its own. A run's citation panel asks the site's tracked questions in the same way. We keep each answer, the engine's full response, the brands it names or cites, the sources it cites and the searches it says it ran, in the workspace, for the life of the site. The searches engines ran are also looked up in public search results (Google's and Bing's through DataForSEO, Brave's for the searches Claude ran), and those searches, with the topics' names and questions, are looked up for their monthly search volume; those public results and volumes are kept once for every workspace, with nothing about who asked.

Shared answers to public questions. Most questions are public: the same question in the same market is the same question for everyone who asks it. When any workspace asked an identical public question of the same engine in the last six hours, the poll reuses that answer instead of asking again. The shared copy holds the answer and nothing about who asked, is deleted within a day, and no workspace can see which other workspace asked. A private question never reads or writes a shared answer: a tracked prompt not marked shareable, and every question taken from the site's Search Console, which is always private. A workspace admin can turn sharing off for the whole workspace in its settings.

Engine Briefs. Once a week, Vercatus asks each engine what it knows about the site's brand, whether it would recommend it for the site's main topic, what would make it the first choice, and its weaknesses and why people choose its leading competitor, with search on and, where the engine allows, off. We keep the answers, their sources, what they say about the brand and a summary per engine, and compare the facts they state with the site's own pages. They stay in the workspace for the life of the site.

Answers and briefs can name people the engines chose to mention, such as a company's founder or an author. We use them only for the customer's measurement and do not extract people from them.

The pages engines cite. When an answer cites a public page of another website, Vercatus fetches that page as itself: under its own name, honouring the site's robots.txt and keeping to its crawl delay, never under another crawler's identity. Some social networks, forums and video sites are never fetched, and a page robots.txt keeps from us is counted as cited without being read. From each page we keep its address, title and kind, its readable text and the brands it names, in their order and rank, as a new version each time the text changes, so that a page's history can be compared; we do not keep the raw page. A page is read again only when answers cite it again. The readable text of each version is deleted 24 months after we last read that version; its address, title, kind and the brands it named stay, so that the dates of a page's changes can still be read. These pages are kept once for every workspace, with nothing about which workspace's answers cited them, so they are not deleted when a workspace closes.

A cited page may contain personal data it publishes, such as an author's name or a quoted reviewer. For these pages we are the controller, on the basis of our legitimate interest in knowing which public pages answer engines rely on. We read them for the brands and facts they state, do not extract people into a directory, do not enrich them and do not use them to train models. If you are named on such a page you can object or ask us to delete what we hold about you, at hello@vercatus.com; a site owner can keep our crawler out with robots.txt, as the crawler page explains.

What we learn across customers

Vercatus learns from what happens across all workspaces, in aggregate only, in three ways.

• What plays did. When a play's verdict is final, its result joins a pool of plays of the same kind, by engine, industry and market, which sets the expected gain of the next such play for every customer. Only final verdicts count, each site counts a limited number in any period so that no one site dominates a pool, and a pool is used or shown only once at least three sites contribute to it.
• Natural experiments. When a public page engines cite changes, for example by starting or stopping to name a brand, and the answers move afterwards, the result is recorded with the page, the brand, the market, the engines and the estimate, and nothing about the workspace whose poll found it. A workspace sees only the experiments its own poll found; otherwise they count only inside the pools above.
• Calibrating AI demand. Where a site has connected its crawler traffic, or added its AI reports from Search Console or Bing, the AI exposure observed on the pages that serve a topic is set against the modelled demand, and these ratios are pooled by engine, market and kind of topic to correct the AI Demand Index for every customer. A pooled ratio is applied only once at least three sites contribute. Analytics sessions are shown beside a site's own estimate and are never pooled, and nothing read through the Google connections is used for this calibration.

The answers a verdict or a natural experiment is read from can include answers to questions taken from a site's Search Console; no search, page or figure read through the Search Console connection, and nothing from Google Analytics, enters a pool. The pools hold counts and estimates only, with no workspace, site, topic, question or person in them. They are never used to show one customer's data to another, and they are kept after a workspace closes because they identify none. We publish learnings on vercatus.com only when we choose to, and then only pools above the same floor.

Data from your Google account

Vercatus reads Google user data in three cases, each started by you, and only reads it; it never creates, changes or deletes anything in your Google account.

• Continue with Google. When you sign in with Google we receive your name, email address and Google account identifier, and use them only to create and sign in to your Vercatus account.
• Google Analytics (read-only scope). When you connect Analytics to a site we list the properties your Google account can read, with their web stream addresses, so you can pick the one for the site. We then read daily aggregate reports for that property: sessions, engaged sessions, key events and users by traffic source, medium and landing page, going back thirteen months. These are totals; no report we request identifies a visitor.
• Google Search Console (read-only scope). When you connect Search Console we list the properties your Google account can read and your permission on each, which is also how the connection can verify that you control the site. We then read search queries and pages with their clicks, impressions and average position over the last 28 days, which queries led to which pages, the daily impressions and clicks of searches for the site's brand, and Google's index status of the site's pages.

With either connection we also receive the email address of the Google account used, so the workspace shows which account is connected.

How we use Search Console data. For the features of your site's workspace, in five ways:

• Your reports. The search tables, summaries and branded-search figures, the index status of your pages, alerts and exports.
• Your topics. Each search is classified by who is asking. When your topics are built, the searches classified as buyer, problem or brand searches, with their impressions, are sent to the language-model provider below (Anthropic) under terms that forbid training on them, so that your topics follow what your buyers search. Other searches (support, documentation, article readers', operators', the echoes of our own checks, and unclear ones) never become topics or questions.
• Your measurement. Those buyer, problem and brand searches join your topics as questions (at most a quarter of a topic set's questions); in a run they also join its question set and the tracked questions its citation panel asks. Each is private to your workspace: the poll asks it of the answer engines as written, every day, to measure your Answer Share; its answers are never shared with another workspace, never taken from another workspace's, and not recorded in the repository we learn from across customers. Asking means the search's words reach the providers of those engines: Anthropic (Claude), OpenAI (ChatGPT, through its API or through its app via DataForSEO), Google (Gemini, and AI Overviews and AI Mode via DataForSEO), Microsoft (Copilot Search, via DataForSEO), Perplexity, xAI (Grok) and Mistral (Le Chat), and You.com where it is switched on for a run's panel. Each receives the question as anyone asking it would, under its own API terms, with nothing about you or your account. As for every question, the language model below reads each answer with its question for the brands and sources it names, and a play's guide can quote the questions of its topic.
• Demand. The monthly search volume of those searches may be looked up through DataForSEO (Google Ads data), to estimate the demand on your topics; the volumes are kept in a lookup table with no workspace on it.
• The analysis of your pages. In a run, each sampled page carries its clicks, impressions, position and top searches; the search checks use them (a page shown often and rarely opened, for example); the run's questions taken from your searches are tested against your pages by the language model, like its other questions; and a summary of your top searches and pages is part of what the language model reads to write the run's analysis.

How we use Analytics data. For the features of your site's workspace: the charts and reports of what AI assistants and search sent the site, the outcomes in Results (AI sessions before and after), whether a fix changed them, goals, your site's own AI sessions shown beside its AI demand estimate (never pooled with other sites), the monthly brief, alerts and exports. It is not sent to the answer engines, not used in the analysis of your pages and not given to a language model, except as described for the assistant below.

What it is never used for. Data read from either connection is not sold, not used for advertising, not shared with another workspace, and not used by Vercatus to train or improve any AI model. In the in-app assistant, when you ask about your site, the assistant can read this data, as you could on screen, and pass it to the language model provider below so that it can answer you, under terms that forbid training on it. Our staff do not look at it unless you ask us to for support, or where security or the law requires it. Conversations with the assistant stay private to your workspace; our staff read one only when you hand it to our team (with "Talk to a person", or by asking the assistant to pass it on), including any figures it quoted, and each reading is recorded in your workspace's audit log.

Who sees it. The members of your workspace, and anyone you choose to send a report, alert or read-only link to. It is stored with the providers that host Vercatus, listed below; the providers named above receive what is described there; it is not shared with anyone else unless the law requires it or Vercatus is sold or merged, in which case we tell you first.

How we protect it. The token that lets us read your data is encrypted (AES-256-GCM) before it is stored and is used only by our servers. The data is kept in the workspace's rows, separated from other workspaces by row-level security, and travels encrypted.

How long we keep it, and how to remove it. The periods are in the table above. Disconnecting Search Console in the workspace deletes, straight away: the token; the search rows (queries, pages and which queries led to which pages); the daily brand-search figures; Google's index status of your pages; the questions taken from your searches, with every answer to them and each answer's stored full response, which is then removed from storage; the tracked questions taken from your searches with their panel results, and the impressions and weights other tracked questions had from Search Console; a run's questions taken from your searches; the search figures attached to your pages in your runs; the Search Console section of each run's analysis input; the run log entries that quoted searches; the top search quoted in a finding's evidence; and the demand estimates that used your searches, with their looked-up volumes. Disconnecting Analytics deletes the token and the daily aggregates. Neither can recall what has already left or been written: what a model wrote from the data before (a finding, a play's guide, an answer of the assistant), reports, emails and exports already sent, figures already frozen (a final verdict, a starting point, a checkpoint as sent), and the questions the answer engines' providers and DataForSEO received when they were asked. Closing the workspace deletes all of it within thirty days. You can also withdraw Vercatus's access in your Google account settings; that stops any further reading, and disconnecting in the workspace, or writing to us, deletes what was already read.

Vercatus's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.

Who processes it for us

These providers process data on our behalf, under contracts that bind them to our instructions and to confidentiality. The list is the same one the data processing agreement carries; we update both together and tell workspace owners thirty days before adding a provider.

Payments are taken by Stripe as merchant of record through its Link service. Stripe decides how it processes the payment, the receipt and any refund or dispute, so for that data it is a controller in its own right under Stripe's privacy policy; we receive the subscription's state, the billing contact and the invoice records, and never the card.

Enterprise workspaces can route analysis through their own model-provider account (Anthropic, Amazon Bedrock or Google Vertex AI); Anthropic is then not our sub-processor for that workspace's analysis, though the poll still asks Claude as one of the answer engines.

Two kinds of service are not on the list because they receive no personal data from us or act on your own account. Public lookups: Wikidata (the Wikimedia Foundation) for the names of brands and the domains they own, and IndexNow when you ask us to tell search engines that a page changed, with its address. Services you connect under your own account with them: Bing Webmaster Tools, your Cloudflare zone for crawler traffic, and the ticketing, chat and webhook destinations you choose for hand-offs and alerts; each receives what you send it, under your agreement with that provider.

Transfers outside the EU

The application and the database run on servers in Amsterdam, Netherlands, so workspace data is stored in the EU. Mistral AI and DataForSEO are in the EU. Other providers above are in the United States: the model providers and the other answer engines, Brave, error reporting, email delivery, the payment processor, and Cloudflare's network, whose object storage holds fetched page bodies for thirty days and the answer engines' full responses for as long as the answers they belong to. We transfer data to them under the EU-US Data Privacy Framework where the provider is certified, and otherwise under the European Commission's standard contractual clauses, with the supplementary measures the provider documents. Keeping every step, including analysis, inside the EU is planned and is stated as planned on the security page; until it ships, the locations above are where data is processed.

Retention

Vercatus retains fetched page bodies for 30 days and then deletes them; answers, briefs, plays, verdicts and reports are kept for the life of the site and deleted within 30 days of the workspace's closure; a cited page's text is deleted 24 months after Vercatus last read it; pooled figures identify no workspace and are kept; account data is kept until the account is deleted; server logs for at most 30 days.

Billing records are the exception: accounting law in Hungary requires invoices and the records behind them to be kept for eight years. A workspace that cancels its plan stays readable for ninety days and is then closed; deletion follows within thirty days of closure.

The page bodies in that statement are the pages of the customer's own site. The poll's answers, the Engine Briefs, the plays and the verdicts are derived data of the workspace and follow the same rule; the answers to questions taken from Search Console are also deleted when it is disconnected. Two kinds of data are kept apart from any workspace and so outlive it: the public pages engines cite, of which we keep each version and never the raw page, and whose readable text goes 24 months after we last read that version, the address, title, kind and brands staying; and the figures pooled across customers, which identify no workspace. Anyone named on a cited page can ask us to delete what we hold about them.

Security

Every record carries the workspace it belongs to and every request is scoped to the caller's workspace; behind that, the database enforces row-level security. Passwords are stored as salted hashes and API keys as one-way hashes, shown once. Connections are encrypted in transit. The audit log records every decision, run, export and setting change. What is in place and what is planned is kept current on the security page; nothing is described there as shipped before it is.

If a breach affects your personal data we tell the affected workspace owners without undue delay and the supervisory authority within seventy-two hours of becoming aware, as the GDPR requires.

Your rights

If you are in the EU or EEA, the United Kingdom or Switzerland, you can ask us for access to the personal data we hold about you, for its correction or deletion, for a copy in a portable format, to restrict or object to processing based on our legitimate interests, and to withdraw a consent you gave. If you are a California resident, you have the corresponding rights under the CCPA and CPRA, including the right to know, to delete, to correct and not to be discriminated against for exercising them; we do not sell or share personal information as those laws define it. Residents of other US states with privacy laws have similar rights.

Write to hello@vercatus.com from the address on your account, or from any address with enough detail for us to identify you. We answer within one month, and tell you if we need longer. Requests about a workspace's crawled content or its poll's answers go to the workspace owner as controller; we help them answer. Requests about a public page engines cite, or about the figures we pool, come to us, since we are the controller for those.

Cookies and browser storage

The workspace sets one cookie, ca_session, which keeps you signed in for thirty days and is strictly necessary; it is not used for anything else and no consent banner is needed for it. The workspace also keeps your display preferences (theme, whether the navigation is collapsed) in your browser's local storage, which never leaves your device. vercatus.com sets no cookies at all. It counts page views with Cloudflare Web Analytics, which sets no cookie, stores nothing in your browser and does not identify you.

Changes

When this policy changes, the version number and date at the top change with it. A change that affects what we collect or who processes it is emailed to workspace owners before it takes effect.

Version 7 describes the measurement as it is built now. It adds: the daily poll of the answer engines and the run's citation panel, with the shared answers to public questions; the Engine Briefs; the public pages engines cite, read under our own name and kept apart from any workspace, with their text deleted 24 months after we last read it; what we learn across customers, in aggregate only; and the use of Search Console searches to build your topics and, as questions private to your workspace, to measure your Answer Share, with what disconnecting deletes. The list of who processes data for us now names each answer engine's provider (OpenAI, Google, Perplexity, xAI, Mistral AI, and You.com where it is switched on), DataForSEO, Brave and Sentry. The retention statement covers the answers, the cited pages and the pooled figures.

Contact and complaints

Vercatus, Budapest, Hungary, hello@vercatus.com. If you think we have handled your data wrongly, tell us first and we will try to put it right. You can also complain to your supervisory authority; ours is the Hungarian National Authority for Data Protection and Freedom of Information (NAIH).