Changelog

VeracityAPI is shipping.

Public product updates. Subscribe via Atom feed.

2026-05-25

Shipped /ai-text-reviewer-subagent — a positioning lander that frames VeracityAPI as the adversarial reviewer subagent in Claude Agent SDK, OpenAI Agents SDK, and LangGraph loops. Includes a bounded-revise pattern explainer and four code blocks (Claude Agent SDK tool, OpenAI Agents SDK guardrail, LangGraph reviewer node, vanilla bounded loop). Wired into the homepage agent-positioning section, the footer Detection APIs group, core + category internal-link sets, sitemap, llms.txt, and agents.json (new reviewer_subagent URL + adversarial_reviewer_subagent capability).

Retired audio + video from the marketing surface. The product now promotes text + image only. /ai-audio-detection-api, /ai-video-detection-api, and related modality-specific landing pages 301-redirect to /what-we-detect. Nine audio use cases retire (e.g., audio-phone-snippet-triage) and 301-redirect to /use-cases. Homepage drops the audio + video demos, modality cards, and pricing line items. Pricing, docs, methodology, how-it-works, /what-we-detect, /for-agents, MCP, llms.txt, agents.json, and OG SVG all scrubbed. Live API endpoints (/v1/analyze-audio, /v1/analyze-video, /demo/analyze-audio) stay alive for back-compat — they're marked deprecated: true in the OpenAPI spec but continue to work for existing customers.
Retired the /vs namespace. The five vs-only content blocks (where-each-wins, modality coverage, output design, pricing notes, migration notes) merged into /alternatives/{slug}; every /vs/{competitor} URL 301-redirects to /alternatives/{alternativeSlug}; /vs/synthid relocates to /integrations/synthid where the layered-integration framing fits naturally.

Site audit fixes: corrected JSON-LD highPrice to $0.05 (was 10× overstated); added video to homepage schema example; cleaned llms.txt type enum and added video pricing line.
Removed /vs and /vs/* from sitemap.xml to stop sending contradictory signals while X-Robots-Tag: noindex, follow remains in force until the 2026 benchmark freeze.
Shipped Atom feeds at /blog.atom and /changelog.atom with <link rel="alternate" type="application/atom+xml"> auto-discovery on every page.
Added FAQPage JSON-LD to /pricing and /docs; switched use-case pages from HowTo to TechArticle schema (Google rejects HowTo without a step array).
Subprocessors page now lists Modal, GitHub/npm/PyPI, and the Anthropic vision use; contact email standardized to hello@veracityapi.com for privacy/deletion requests and security@veracityapi.com for security disclosures.
Brand consistency: /privacy title corrected to VeracityAPI Privacy; /.well-known/agents.json now linked via rel=alternate in every page head.

Shipped installable SDK packages: @veracityapi/sdk for TypeScript and veracityapi for Python.
Updated docs, examples, and homepage copy around SDK-first integration paths.
Repositioned homepage around content trust infrastructure for AI products while keeping agent-native proof points.
Simplified pricing around self-serve prepaid credits and custom-volume contact paths.
Added public /status and /changelog surfaces.
Promoted MCP as a primary agent-distribution path.
Reworked README into a developer landing page with curl, TypeScript, Python, MCP, pricing, privacy, and routing eval evidence.

Published seed text eval artifacts: 500 samples, 88.0% routing-action agreement, macro F1 0.871.
Added transparent eval caveats: routing-action quality, not forensic authorship evidence.
Added unified POST /v1/analyze support across text and image URLs. Audio/video typed endpoint experiments were later retained as deprecated back-compat-only routes.

Added audio URL workflow triage under billing bucket audio_v0 with transcript return.
Added privacy-safe audio logging: URL hash/hostname only, no raw audio bytes, no base64, no full URL storage.

Added image URL analysis with synthetic_image_risk, evidence, limitations, and action routing.
Added public image demo fixture and image-specific privacy docs.

Added /llms.txt, /.well-known/agents.json, /openapi.json, sitemap, and robots metadata.
Added /for-agents with routing policy templates.

Added action-first response contract: allow, revise, human_review, and reject.
Added derived trust fields and privacy defaults centered on store_content=false.

Internal links

Map text slop, image forgery, and workflow-risk signals to routing actions.

Wire VeracityAPI in as the adversarial reviewer in Claude Agent SDK, OpenAI Agents SDK, and LangGraph loops.

Production retry, billing, auth, and validation behavior for agent integrations.

Drop-in wrappers for LangChain, queues, cron jobs, and moderation pipelines.