What changed, and what it enables.

Every visibility conclusion now shows how much trust it has earned

• Executive reports distinguish Not collected, Early signal, Directional evidence and Repeatable evidence.
• Valid observations, citation-capable evidence, exact-scope remeasurements and automated completion are shown together.
• Failed, empty, truncated and provenance-incomplete runs remain visible as excluded attempts but never become negative observations.
• The HowAICite Agent returns the same evidence maturity and limitations through its visibility snapshot.
• The public methodology publishes the thresholds and states that evidence maturity is not market share, search volume or a statistical confidence interval.

Read the Evidence Contract →

Clearer audit outcomes and a controlled measurement note

• Audit reports now lead with the bottom line, separate supported conclusions from unsupported claims, and prioritize next actions.
• Passed technical checks collapse so unresolved work stays visible.
• Public sample reports now show report IDs, declared scope, collection surface and what was not verified.
• Published an aggregate Australian HR software measurement note with failures, denominators, competitor counts and explicit provider-API limitations.
• No user session or private workspace evidence is included in the public note.

Read the controlled measurement note →

Security, support and provider-API scope made explicit

• Published a security and data-handling page with current controls, retention, service providers and unavailable enterprise features.
• Added an operational first-response target for support without presenting it as a contractual SLA.
• Clarified that automated quotas cover provider-API observations, not consumer ChatGPT, Google AI Mode, Google AI Overviews or Copilot collection.
• Removed “Planned” public API and white-label entitlements from self-serve plans; either capability requires an explicit written HowAI Suite scope.
• Linked the public research repository directly from Research and Methodology.

Review security and data handling →

OpenAI, Claude, Perplexity and Gemini in one evidence workflow

• Saved buyer questions can now run on demand across all four configured providers.
• OpenAI, Claude and Perplexity preserve provider-reported citation URLs with each answer.
• Gemini preserves model-answer mention evidence, token usage, model and latency as a separate non-grounded surface.
• Every completed run can feed Why Not Cited, competitor gaps, opportunity work and Before → After rechecks.
• Workspace quotas and provider health are visible before a team starts a measurement.

Review current provider coverage →

Gemini API connector with an explicit no-web boundary

• Non-grounded Gemini API responses can be collected as their own automated surface.
• Answer, response ID, model version, token counts, latency and requested scope are preserved.
• No Google Search Grounding tool is enabled and no citation URLs are claimed.
• Results are labeled as model-mention evidence, separate from Gemini consumer, Google AI Mode and Google AI Overviews.
• Country, region and city are retained as requested scope but are not claimed as provider-applied location.

This connector does not measure Gemini consumer search or Google citation visibility. Google Search Grounding remains disabled for longitudinal analysis under the standard grounding terms.

Citation evidence, without search-context inflation

• Provider-reported citations are counted separately from retrieved search results and links typed in answer prose.
• A cited page without a brand-name mention is recorded as its own outcome. A source link is not treated as a recommendation.
• Incomplete answers retain their diagnostic output but do not enter successful visibility observations.
• Model, parser and collection basis changes create separate comparison cohorts.
• Atomic quota checks, safe request replay and rate-limit retry delays protect measurement activity.
• Pending runs refresh automatically; completed runs link directly to their evidence.

Configured credentials do not prove successful collection. These changes improve measurement integrity; they do not establish statistical accuracy or consumer-chatbot equivalence.

Perplexity Sonar API connector

• Perplexity Sonar Pro added as an independent automated evidence surface.
• Country, region, city and supported language preferences are passed to web search.
• Answer, citation URLs, search-result titles, model, tokens, latency and provider cost metadata are preserved.
• Perplexity API observations remain explicitly separate from the Perplexity consumer interface.

Credential readiness does not prove a successful paid request. A controlled measurement is still required before operational activation is claimed.

OpenAI and Claude Web Search API connectors

• Independent server-only credentials, model settings and readiness states for each provider.
• Per-provider run controls with a shared workspace quota.
• Answer, citation source, model, token, latency and failure provenance preserved for every run.
• Provider API observations remain explicitly separate from consumer ChatGPT and Claude interfaces.
• Autopilot schedules only selected connectors that are configured and within allowance.

A configured key is not evidence of a successful run. Each provider still requires available credit, and every run records its own outcome.

Public product evidence and research foundations

• About, Product and Coverage pages with explicit availability.
• Documentation distinguishing readiness, mentions, recommendations and source links.
• Public historical audits of HowAICite pages with downloadable raw findings.
• A frozen 30-question GEO evaluation panel and a recommendation–citation study protocol.
• Downloadable research validation and summary tools, with empty data kept distinct from zero visibility.
• Consistent product identity, article metadata and social previews.
• Pricing separates proposed packages from features available today.

Explore the research materials →

Website audits and private evidence workspaces

At initial release, readiness checks, account access, saved prompts, manual AI answers, source analysis, improvement plans and an intelligence ledger were available.

This historical entry has been superseded by the dated releases above. Consult current coverage rather than using this entry to infer today's provider or collaboration availability. Public API access is not included in standard plans.