Privacy and data use

The public diagnostic works without an account. Here is what the service processes, what stays on your device and what the private owner workspace stores.

Public website crawls and campaign reports

The website check includes a public crawl of up to 20 eligible pages in small batches. You can also run it separately in the crawl workspace. The destination website sees those requests. Page observations, retained text and reports stay in your current browser tab unless you explicitly save an improvement-plan copy on this browser. They are not saved to the private workspace. Continuation tokens expire after 30 minutes and are bound to the requesting client. Downloaded reports contain the inspected website addresses and evidence, so keep them only where you intend to share that information.

Abuse controls retain daily-rotated client digests, session identifiers and request counts, not submitted website addresses or page content. Expired session counters are pruned after approximately 60–90 minutes, client/hour counters after approximately 48–49 hours, and global daily counts after approximately seven to eight days. Cleanup runs during crawl requests and scheduled maintenance; it depends on the service completing those jobs.

GEO campaign setup and analysis run in the browser. Imported AI answers, collection worksheets and earlier campaign reports are processed locally and are not sent to model providers. Campaign exports contain source answers, prompts, supplied settings and observation times. A downloaded approved execution bundle contains the original and proposed HTML, source evidence and approval information; treat that file as private project material. Running its local check, apply or rollback commands does not upload it or publish the website.

Public website diagnostics

When you submit a URL, SingCompare’s server receives it and fetches the submitted public page, selected discovery files and up to three additional eligible pages on the same origin to generate the initial report. The homepage then adds the bounded public crawl, Google performance checks and browser-local plan described below to one combined report. Discovered pages are subject to crawler rules and strict request limits. The destination website can see the resulting server request. Only submit public addresses; do not include private tokens, account links or confidential query parameters.

The public diagnostic does not save the full target report as a private account history. You can download the complete PDF or JSON result yourself. Aggregate service records describe use, completion, failure and duration without storing the submitted target URL in analytics. Operational infrastructure may process request data as described below.

The two-page comparison sends each submitted URL to the same public diagnostic service, one at a time. Each fresh check uses the normal diagnostic quota. Both reports and their comparison stay in the current tab and are not saved as account history or in browser storage. Downloads stay on your device until you delete them.

The current public diagnostic does not send the page to paid AI-model APIs. Its GEO result concerns technical readiness, not a live answer from an AI provider.

Comparing an earlier diagnostic JSON happens in the current browser tab. The selected file is not uploaded or saved in browser storage. The baseline remains user-supplied and unverified. Clear the comparison or run a new diagnostic to remove it from the active view; downloaded files remain on your device until you delete them.

Public Intelligence preview

After a public website diagnostic, the website check automatically prepares an Intelligence plan from that diagnostic in your browser, using a medium effort allowance. This planning step prioritises actions, compares approaches and explains evidence gaps without a model request or access to private owner records. The accompanying Google performance checks and deeper crawl make their own requests, described in this notice; their findings are separate sections of the same report.

The preview and its source diagnostic remain in the current tab. They are not saved as an account history. A starting-point copy can be retained through the optional saved improvement plan described below. Starting another diagnostic, reloading or leaving removes the working preview; an explicitly saved starting-point copy remains until deleted. Downloaded PDF and evidence JSON files remain on your device until you delete them. The public preview is separate from the retained private Intelligence briefs described below.

Public name discovery

When you choose Check public presence, SingCompare receives the name, entity type and optional context you enter. The server searches Wikidata and GDELT news metadata using the name; the optional context also narrows the news query. These sources receive the search terms and server request under their own service and privacy arrangements. This pilot does not contact AI models, collect review-platform ratings or fetch the linked news articles.

The application does not store your search terms or discovery report. Results and your profile selection stay in the current tab until you clear them, reload or leave. PDF and JSON exports remain on your device until you delete them. Network infrastructure and the source providers may process request information; this is not a promise of zero third-party logging.

Separate abuse-prevention records contain daily-salted network-address hashes and aggregate request counts. Fixed provider cooldown records contain a provider identifier and expiry time, not your search. These records limit requests and respect source backoff instructions. The optional manual-feedback workspace below has separate device-saving controls.

Wikidata structured data is available under CC0. News metadata is attributed to GDELT; linked publishers retain their content rights. Directory records and matching headlines can be incomplete or concern someone else with the same name. A selected profile does not independently establish identity.

Browser-only analysis tools

The public Search Data tools read CSV files in your current browser tab. Query performance, linking-site summaries, linked-page summaries and individual-link exports are not uploaded to SingCompare, sent to AI services or saved in browser storage by these tools. They remain in memory until you clear them, leave the page or close the tab. Downloads remain on your device until you remove them.

The Performance tool reads Lighthouse JSON locally. The optional manual-feedback workspace reads supplied feedback, entity details, review ratings and case notes in the current tab, without requiring a website. Neither public tool uploads that material or sends it to AI services. Performance data stays in memory. Reputation saving is off by default: you may explicitly save a copy on this device using browser storage, then choose whether to restore it on a later visit. Anyone using the same browser profile may access that copy. Disable saving or use the saved-copy delete control to remove it; clearing this site’s browser data also removes it. Browser storage may be cleared or become unavailable, so export JSON for a copy you control. Without saving, closing or reloading removes the working copy. Only include information you have a right to use, and avoid unnecessary personal details.

Use data you are authorised to analyse. Export labels, dates and property settings are supplied by you; importing a file does not independently verify its source, current search position or live backlinks.

Page after loading

Submitting the homepage check or rechecking a page can send its public HTTPS address to Jina Reader for one browser visit. SingCompare requests a fresh, uncached result and asks the service to respect crawler permissions. The request contains no account credentials, uploads or visitor cookies. Selected short passages appear in the same report. Region and page versions can affect the comparison.

Optional Google PageSpeed test

When you submit the homepage website check, SingCompare sends the public HTTPS page URL to Google PageSpeed Insights for a mobile test. You can also choose Test this page in Performance with your preferred device. Google visits the page to run a Lighthouse lab test. Do not submit private or secret-bearing URLs. Query parameters, fragments and credentials are not accepted. Google processes requests under its own privacy practices.

The application does not save submitted URLs or test reports. Results stay in your tab unless downloaded. Temporary daily salted network-address hashes and aggregate counters enforce limits; raw addresses are not stored in these records. Counters are removed after two UTC day boundaries. No test runs automatically on page load, and uploaded Lighthouse files remain local.

Optional Chrome visitor-data lookup

When you submit the homepage website check, SingCompare sends the public website origin and mobile device type to Google’s Chrome UX Report API. You can also choose Look up field data in Performance with your preferred device type. The application does not persist the submitted origin or returned report. Results remain in the current tab unless you download them; reloading removes the working copy. Lighthouse files are not uploaded. Google processes this request under its own privacy practices.

Daily salted network-address hashes and aggregate quota counters limit this lookup. No raw network address is stored in these application records. Hourly maintenance removes old counters after two UTC day boundaries. No lookup runs automatically on page load.

Website analytics and performance

SingCompare records limited activity to understand which public pages and diagnostic actions are useful and whether the service works. Records contain a UTC day, an allowed public page path, an event name, a broad referral-source category, a broad device category and aggregate counts or durations.

This application does not store a full referrer, URL query string, submitted diagnostic target, persistent visitor identifier or raw IP address in these analytics records. It does not set analytics cookies, fingerprint browsers or build personal browsing sessions. Aggregate analytics are retained for up to 90 days.

On selected public information pages, the application also receives numeric Largest Contentful Paint (LCP), Interaction to Next Paint (INP) and Cumulative Layout Shift (CLS) measurements, a fixed page path and broad device category. These describe visits to SingCompare, not the websites checked by visitors. A hashed, temporary measurement identifier prevents duplicate samples; it is not a persistent visitor identifier. Samples expire after 30 UTC day buckets through hourly maintenance. The collector excludes the private admin and search-data, performance and reputation workspaces, omits URLs with query strings and referrers from its payload, and respects the same privacy signals.

Browser events can be incomplete or repeated and do not identify unique visitors. Server-confirmed diagnostic metrics are recorded separately from browser events. The application honours Global Privacy Control and Do Not Track signals for telemetry; this does not disable essential security checks needed to operate the service.

Abuse prevention and infrastructure

An incoming request necessarily exposes network information to the hosting infrastructure. SingCompare uses Cloudflare to serve the website and run its backend. Cloudflare may process request metadata under its own service and privacy arrangements. This notice describes application storage; it does not claim that no network data is ever processed or logged by infrastructure.

For abuse prevention, the application derives short-lived rate-limit keys from network addresses using a daily salt. Raw addresses are not persisted in these application rate-limit records. The keys are scheduled for deletion after 48 hours; hourly maintenance can delay removal beyond that threshold. Separate aggregate diagnostic quota counters retain up to seven UTC day buckets. These controls limit abuse and costs without creating a long-lived visitor profile.

If Cloudflare Turnstile is enabled, a managed browser challenge processes network and browser information through Cloudflare to check for automated abuse. The server verifies its short-lived token before starting a diagnostic. Turnstile is optional and is not active until configured. See Cloudflare’s Turnstile privacy information.

Do not use the diagnostic or evidence fields to submit secrets or personal records. The service inspects public website information and is not a secure document vault.

Optional AI evidence notebook

The separate batch-comparison tool keeps imported AI observations in memory only. These batches are not uploaded or added to persistent notebook storage. Clearing the comparison, reloading or leaving the page removes its working copy; downloads remain on your device.

The optional public notebook handles information you enter or import in your browser, including brand settings, questions and AI answer text. Saving the notebook on the device is opt-in and uses browser storage. Without that choice, do not assume the notebook will remain after closing or reloading the page.

The notebook does not automatically collect answers from external AI services. If you use another service to obtain an answer, that service’s terms and privacy practices apply. Review pasted answers for private information before saving or exporting them.

You can export or clear the notebook using its controls. Clearing browser site data also removes saved local records. An export is a file you control; deleting the notebook does not delete copies you downloaded or shared.

Private admin and Google Search Console data

The owner workspace is protected separately from the public diagnostic. Its secure authentication cookie supports an admin session, not advertising or public visitor analytics. Owner authentication is required to view private measurements and Search Console imports.

When an authorised owner imports a Search Console file or retrieves data through a configured Google connection, the supplied property information and query, page, date and metric rows are stored privately. Upload only data you are authorised to use. The workspace retains up to six imports of up to 200 rows each; an import remains until deleted or displaced by the retention limit.

A Google connection requires explicit authorisation with read-only Search Console access. Connection tokens are encrypted in server-side storage, not placed in public website code. Scheduled retrieval is off by default and requires the owner to enable it. Disconnecting removes saved connection tokens, pending authorisations and the retrieval schedule; delete retained imports separately if you also want their stored data removed. You can additionally revoke the application’s permission in your Google account.

Search Console data is used to display the owner’s search performance and support analysis of the selected property. It is not exposed to public diagnostic users or sold for advertising.

Period comparisons use the selected stored datasets. CSV filter confirmation is an owner assertion, not independent verification. Comparing or exporting these records does not send them to AI providers.

Private website workflows store an origin-specific ownership challenge and its verification dates. The proof file is deliberately public; it is not an account password. Revoking verification removes its stored record and prevents new approved-change operations until the site is verified again.

When the owner creates a change proposal, SingCompare stores the uploaded static HTML, proposed version, file path, diff, approval and recheck evidence in that private workspace. Files are limited to 64 KiB and the workspace holds at most ten proposals. Do not upload credentials or confidential source files. Proposal contents remain until the owner deletes the proposal or its monitored website; deletion does not remove previously downloaded files.

When a verified owner starts a deeper website crawl, the private workspace stores queued URLs, observed page metadata and status, a bounded internal-link sample, up to 3,000 characters of visible static content per inspected page, crawl progress and findings. It retains at most three crawls per website, with up to 500 page attempts each. Full fetched HTML is not retained by this crawler. Pause stops new foreground batches; an admitted batch may finish. Deleting a crawl or its monitored website removes the corresponding saved crawl records. The target website sees the crawler requests.

Private Intelligence briefs retain a fixed copy of the selected audit, optional compatible Search Console dataset and generated analysis. These copies remain after the original audit or import is deleted. The workspace retains at most five briefs per monitored website, each with up to 1.5 MiB of source snapshots and generated output. Delete the brief or its monitored website to remove these copies; downloaded exports must be deleted separately.

Generating an Intelligence brief analyses existing saved evidence: it makes no new crawl, calls no AI provider and sends no source data to external services. Briefs are private to the owner workspace. The current review status and note, up to 1,000 characters, are retained with the brief and may be changed; previous review notes are not kept as a history. Accepting a brief records a planning decision, not permission to publish a website change.

Improvement-outcome plans retain a fixed hypothesis, query cohort, selected baseline snapshots, application attestation and follow-up measurements. These private copies remain after the original imported dataset or audit is removed, so the plan can preserve its evidence. Delete the outcome plan or monitored website to remove its retained copies. The workspace retains at most ten plans per website. An owner attestation does not independently prove deployment, and an observed metric change does not establish causation.

The authenticated owner may explicitly upload a Reputation workspace JSON to private server storage. This is separate from public name discovery and the optional manual-feedback tool. The private workspace retains at most five records, each with up to 200 feedback rows and 200 cases, with a 1.5 MiB normalised file limit. Records contain entity details, supplied feedback, case notes and resolution attestations. The last five save events retain dates, revision numbers and counts, not earlier content snapshots. Data remains until the owner deletes its record; deletion also removes this save history. Device copies and downloaded files must be deleted separately. The service does not post review responses or send this content to AI providers.

GitHub publishing is disabled until server-side repository access is configured. If enabled, an explicit draft pull-request action sends the approved change and its supporting context to the configured GitHub repository. A public repository makes that material public. Deleting a SingCompare proposal does not delete its GitHub branch or pull request; manage those separately in GitHub.

Optional search API tools

The private owner may separately authorize read-only Google Analytics access. The existing encrypted Google connection then supports on-demand aggregate reports for the SingCompare GA4 property: active users, sessions, views and events for the previous 28 complete Singapore days. Reports and their quota metadata remain in the private tab and clear on sign-out; temporary quota and request counters remain on the server. Disconnecting Google removes the stored Google credentials. This feature does not install a visitor tracking tag, send diagnostic content to Google Analytics, change Analytics settings or make paid AI calls.

The private owner can explicitly read Bing traffic for singcompare.com. This sends the fixed public site URL to Bing Webmaster using a server-side secret. Only bounded daily clicks and impressions are returned to the private tab, with source and retrieval time. Reports are not persisted or automatically refreshed; missing data is not treated as zero. The account API key is never exposed to visitors.

Choosing CrUX history sends the public website origin and device category to Google for up to 25 weekly snapshots. Each snapshot describes an overlapping 28-day collection window. Missing samples remain unavailable; results stay in the current tab unless downloaded.

The private owner may explicitly inspect a URL in an authorised Search Console property. This sends the selected URL and property to Google using the existing read-only connection. The indexed snapshot is displayed temporarily and is not stored as a report. Inspection does not request indexing.

After explicit confirmation, the private owner may notify IndexNow participating search engines about selected public SingCompare pages. The public page URLs, site hostname and public ownership proof are sent to IndexNow. There are no automatic submissions. A receipt is not proof of indexing. This feature cannot submit private pages or other websites.

Saved improvement plans

A starting point and your work-status notes stay in the current tab by default. You can explicitly save one plan on this browser or download a saved-plan JSON file. A device copy includes the diagnostic, available performance and crawl evidence, source URLs and dates, and notes you choose to record. It is limited to 2 MiB and is not encrypted or synchronised to an account. Anyone using the same browser profile may be able to read it.

Device saving is off until selected. While enabled, saving a work status or replacing the starting point updates the device copy. You can turn saving off or delete that copy; deletion does not remove downloaded files or another browser’s copies. Browser settings may also clear device storage. Storage or deletion failures are displayed without claiming success.

Restoring a plan reads it in your browser and does not submit its evidence or notes to SingCompare. Saved and imported evidence is not independently verified. Resuming does not run a new check. Clicking Recheck uses the normal public website requests and quota controls. Work labels describe your own records and do not confirm a fix, deployment or ranking improvement.

Public archive checks

The combined website check sends the submitted public page URL to Common Crawl and Internet Archive to look for existing crawl records and historical captures. It does not create an archive or fetch archived page content. Results contain source names, capture dates and bounded metadata. They remain in the current tab unless you download or explicitly save your plan on this device.

Archive checks share a 20-attempt daily allowance, a one-attempt-per-minute global limit and three attempts per connection/day. Temporary rate counters contain salted daily digests, not raw IP addresses or submitted URLs. Provider outages, rate limits and missing captures do not lower your website score. These records do not prove Google indexing, AI training or AI recommendations.

Your controls and this notice

You can use the public diagnostic without registering, decline persistent notebook storage, clear local evidence, and use browser privacy controls. The private owner can remove imports and disconnect Google access. Public visitors are not enrolled in marketing emails or paid tracking services by using the diagnostic.

This release has no public customer registration or payment flow. Changes that introduce new data uses require an updated notice. The publication date below identifies this version; it does not mean that linked third-party policies share the same date.

Related information: Cloudflare’s privacy policy, Google’s privacy policy and the SingCompare measurement guide.