Terms of Service Generator

Free plans include per-request character ceilings, analyses per hour, and monthly character pools so the service stays responsive for everyone—compare rows on the Pricing page for Starter, Pro, Expert, and Enterprise upgrades. Paid tiers raise or remove those caps, add DeepScan for AI detection where applicable, and expose API keys for automation. Cached identical requests may not re-burn quota when the product implements deduplication, which helps iterative editing. Marketing automation or LMS integrations should budget monthly characters using last month’s logs plus campaign growth. If you spike during a launch, temporary upgrades beat surprise throttles. Overall, limits align with SaaS norms: generous for individuals, scalable for teams.

Inputs are processed to return results; short-lived caching may speed repeat views while ephemeral memory clears. Longer retention—project history, team dashboards, or compliance archives—depends on settings shown in your account and the Privacy Policy. You can usually delete history items from the dashboard when supported; Enterprise customers may configure retention windows or region pinning. GDPR and FERPA-style obligations mean schools should minimize personal data in pasted text and use institutional agreements when required. For regulated workloads, run redacted samples during pilots. Bottom line: assume processing at minimum, confirm retention rules before uploading sensitive drafts.

Pro and Expert plans include REST API access for programmatic detection, readability, humanization, plagiarism, summarization, translation, and related endpoints documented on the API Documentation page. Authentication uses server-side secrets—never embed keys in mobile or browser bundles. Rotate credentials when staff depart and scope separate keys per environment (staging vs. production). Stripe-backed billing records usage for overage invoices when enabled. For batch jobs, backoff on HTTP 429 responses and respect payload size limits published per endpoint. If you need private networking or VPC peering, Enterprise sales can discuss architecture. In conclusion, the API mirrors interactive tools so you can embed SynthQuery wherever content is authored.

SynthQuery continuously retrains and evaluates models on internal corpora and public benchmarks; headline figures for AI detection appear on the Methodology page with scope and limitations spelled out. Independent studies such as Weber-Wulff et al. (2023) on AI-text detectors remind buyers that accuracy varies with text length, topic, and post-editing—no vendor score transfers blindly to every domain. Readability metrics inherit assumptions from mid-century formulas (Flesch, 1948) that work best on prose, not poetry or tables. Always combine automated scores with domain expertise: physicians should review medical claims, attorneys should review contractual language. Use metrics to triage drafts, not to replace institutional judgment. Overall, expect strong lab performance with documented caveats for messy real-world inputs.

Most analysis tools are optimized for English because benchmarks and customer demand concentrate there; UI copy and default tokenization reflect that priority. Translation, grammar, and some detectors expose additional languages—check the per-tool picker for the latest list. Multilingual classrooms and global support teams should run each language separately rather than mixing code-switched text in one paste, which can confuse models. UNESCO highlights growing generative-AI use across languages, so we expand coverage as data quality allows. When a language is unsupported or experimental, results may be unreliable—have fluent reviewers validate output. In summary, start from English excellence, then confirm other locales case by case.