Privacy

Next Marketing Technology Ltd is the data controller for geo.new, registered in England and Wales under number 14384054, at Flat 1, 25 Daleham Gardens, London, England, NW3 5BY. This policy is written under the UK GDPR and the Data Protection Act 2018.

Questions, requests and complaints go to [email protected] and are read by a person.

controller

Next Marketing Technology Ltd

company no.

14384054 (England and Wales)

registered office

Flat 1, 25 Daleham Gardens, London, England, NW3 5BY

contact

[email protected]

Five things, and this is the whole list.

Your address is hashed at the edge of the application, before anything is written down. What the database holds is a keyed HMAC-SHA256 digest used to count requests; the address itself is never stored in it.

the URL

the address you submit, and the report produced from it; its host name and score are listed publicly on /latest, where the newest report for that host can be opened by anyone

a cookie

geo_session — httpOnly, SameSite=Lax, seven days. It is how a report knows it is yours. There is no second cookie.

your IP

kept only as a keyed hash, for rate limits and abuse counters

an API key

if you create one: stored as a hash, with the count of audits it has run

your theme

light or dark, in your own browser’s storage. It is never sent to us.

No name, email address, postal address, phone number or payment details. There is no account to hold them in and nothing to buy with them.

No advertising cookies, no cross-site trackers, no fingerprinting, no data brokers, no marketing list. Nothing here is sold or shared for marketing, at any price.

Audited URLs, query strings, page content and model prompts are deliberately kept out of our application logs.

To run the audit you asked for and to show you the report afterwards. That is the service you requested, and the URL and the session cookie are what make it possible.

To keep the tool standing: our legitimate interest in preventing abuse and staying inside the cost ceilings a free service has. The hashed address and the request counters do that and nothing else.

To know whether it is used at all: our legitimate interest in aggregate counts — audits per day, keyed against keyless. No profile of a person is built, and none could be.

To show the tool at work: our legitimate interest in listing what has been audited lately, which is why /latest publishes a host, a score and that host’s newest report, and nothing that leads back to a person.

Expiry is enforced by the database on a schedule, not applied by hand when someone remembers.

reports

seven days from the run — the report, its evidence, its screenshot and any share made from it

sessions

seven days, and no longer than the reports attached to them

admission records

twenty-four hours (hashed address, hashed target)

request counters

one hour

API keys and usage

while the key exists. Revoke it, or ask us, and it goes.

One script, from TinyAnalytics, counts which pages get opened. It is cookieless and sets no identifier that follows anyone to another site.

It is configured to skip every path that carries an audited URL, to mask report and share addresses to a bucket, and to ignore query strings. The analytics record cannot say what anybody audited, because it is never told.

Where a report shows what ChatGPT answered about a brand, the same script counts a few named events: the answer shown, a question copied, a page audit opened from it, a share created, and a yes or no to "Was this useful?". Each carries only whether the report is private, shared or the sample, and one word for the result (named, not named, not recognized, skipped or failed). None carries the audited URL, the report, the brand or the answer, and each is counted once per report in a browser, remembered in that browser alone.

Google AI Mode’s answer is counted the same way under its own event names, which begin ai_mode_: the answer seen (when it was at least half on screen for a second), a question copied, an "Ask Google AI Mode" link opened, a page audit opened from it, and a yes or no to "Was it useful?". They carry the same two properties and nothing more.

The text extracted from an audited page is sent to Cloudflare’s Workers AI to write the interpretation section of the report. It is sent for that one call. We do not train anything on it and we pass it to nobody else. The ChatGPT questions, where a report asks them, carry only the page’s declared name and its host and, when the page’s content policy permits AI input, a category taken from its title. The Google AI Mode questions carry the same name and host and the category ChatGPT gave in that report.

Four lookups leave our network about the audited brand rather than about you: public entity records at Wikidata and Wikipedia; a search-result probe through scrape.do; two questions to ChatGPT, sent through scrape.do, which passes them to OpenAI; and two questions to Google AI Mode, sent through scrape.do, which passes them to Google. None carries anything about the visitor who started the run. Asking can prompt ChatGPT’s own search to visit the audited host.

Two requests can leave your browser rather than our network. Where a report shows the brands an assistant listed, each listed brand’s icon loads from Google’s favicon service, which receives that brand’s domain and your IP address. It never receives the report, its address or the audited site’s own domain. And if you follow an "Ask Google AI Mode" link, your browser sends Google that question, which names the audited host, under Google’s own terms; nothing is sent unless you click it.

Processing happens on infrastructure inside and outside the United Kingdom. Where a transfer leaves the UK it relies on the standard contractual clauses and the UK addendum in each provider’s terms.

Cloudflare

hosting, the edge network, the job queue, the browser that takes the screenshot, and the model that writes the interpretation

Neon

the PostgreSQL database the reports live in, over verified TLS

scrape.do

the Google brand and review searches, and the two ChatGPT and two Google AI Mode questions a report may ask

TinyAnalytics

cookieless page analytics

Google

the icons of the brands an assistant listed, loaded by your browser from its favicon service; Google AI Mode’s answers, through scrape.do; and the question your browser sends if you follow an "Ask Google AI Mode" link

Any public page can be audited, and pages get audited by people who do not own them. The report then holds that page’s own public content, including whatever personal information its author chose to publish there.

If you operate a page and want a report about it removed, write to [email protected] with the address. It will be deleted — and it would have been deleted within seven days regardless.

To keep GeoNewBot away altogether, disallow it in robots.txt. The audit reads robots.txt first and stops there when it is told to.

You may ask for access to your data, correction of it, erasure, restriction or portability, and you may object to processing we base on legitimate interests.

There is no account here, so identifying “your” data usually means telling us a report id or a share token. Ask at [email protected]; the answer comes within a month, normally much sooner.

You can also complain to the Information Commissioner’s Office, the UK regulator, at ico.org.uk. We would rather you told us first and gave us the chance to fix it.

Session tokens and share tokens are stored as hashes and never in the clear. Reports are served no-store and are excluded from indexing. The database connection uses certificate-verified TLS.

The code that touches raw page HTML runs isolated from everything else and may return only a bounded, structured summary, so a hostile page has nothing to reach.

No system is airtight. What this one offers is a small surface and a short retention: little is collected, and what is collected does not stay.

This is a developer tool and it is not directed at children. We do not knowingly collect anything from a child under thirteen.

The date at the top of this page is the date of this version. If what we do with data changes, this page changes first.