AI Usage & Disclaimer Policy

• AI-generated output — text, explanations, recommendations, or content drafts — may be incomplete, out of date, or simply wrong. It is a starting point, not a verified fact.
• We do not guarantee the accuracy, completeness, or fitness for any particular purpose of AI-generated output.
• You remain solely responsible for reviewing AI output before relying on it, and for any action you take (or don't take) based on it — especially for decisions with real financial, legal, or business consequences (e.g. publishing AI-drafted content publicly, or making a spending decision based on an AI-generated recommendation).
• AI features are not a substitute for professional legal, financial, tax, medical, or other licensed advice.

Where Groager's AI is used to explain a metric, score, or recommendation, our internal engineering principle (which this policy reflects honestly, not aspirationally) is that the AI explains, it does not decide facts. Concretely: our backend calculates numbers, evidence, and confidence levels using deterministic business logic first; the AI's role in those cases is limited to summarizing or phrasing that already-computed evidence in natural language — it is not permitted to invent a number, a data point, or a piece of evidence that isn't already present in what the backend gives it. Where the backend cannot calculate something with sufficient confidence from real evidence, the product is designed to say so honestly (“insufficient data”) rather than have the AI estimate a plausible-sounding placeholder.

The following features send relevant data to Anthropic's Claude API in real time and return a genuine model-generated response (see our Privacy Policy, Section 7, for what data is sent):

• AI Assistant (chat): the conversational assistant in the dashboard is a live, streaming Claude conversation, grounded in your account's real structured data where relevant.
• AI Writer (Content Intelligence): blog draft generation, meta description generation, page title suggestions, FAQ generation, content idea generation, and content optimization are each a real Claude call generating original text from your prompt/inputs.
• “Fix with AI” / AI Recommended Fixes (Website Intelligence): the plain-language explanation, fix steps, and estimated time for a given SEO/technical issue are generated by Claude, given the real, already-detected issue as input — Claude explains a real finding, it does not invent which issues exist.
• AI Overview Ranks: checking whether your domain appears in Google's AI-generated search overviews for your target keywords combines real web scraping with a Claude analysis pass over the scraped result to determine citation.
• Weekly/Executive AI Summary: the narrative executive-summary text (headline, highlights, concerns, top opportunity) is generated by Claude from a real, computed snapshot of your traffic, SEO, and performance metrics for the period — Claude writes the narrative, it does not compute the underlying metrics.
• Revenue Intelligence — explanation text only: where Revenue Intelligence shows a risk or metric, the surrounding plain-language explanation is generated by Claude from the already-computed evidence and confidence result; the risk calculation itself (see Section 4) is not AI-generated.
• Social Post Generator: draft social media post copy for a topic/platform/tone you specify is generated by Claude.

To be transparent rather than let a UI label overstate what's happening: the following are not live calls to Claude or any language model. They are ordinary, deterministic calculations and templated text, computed the same way every time from the same inputs — no model inference involved, even though the product surfaces them under an “AI”-branded name for consistency with the rest of the platform:

• Growth Score / Website Health Score: a weighted, formula-based composite of real underlying metrics — not model-generated.
• Revenue Intelligence's underlying risk/metric calculation: the number and its confidence level are produced by deterministic business logic (evidence resolution → calculation → confidence gate) — only the accompanying explanation sentence (Section 3, above) is AI-generated.
• Growth Center's Risks & Opportunities and growth-summary coach text: derived from real threshold-based rules applied to your Growth Timeline data, and templated accordingly — not generated by a live model call.
• Business Goals progress tracking: plain arithmetic against goals you define — not AI-generated.
• Website Intelligence's “AI Executive Summary” headline text and “AI Score”: generated by a deterministic rule-based template/formula from your real audit data, not a live Claude call for that specific text (this is distinct from the per-issue “Fix with AI” feature on the same page, which is real — see Section 3).
• Today's Priorities' “AI Decision Summary”: a rule-based selection and summary of your real, computed priority signals — not model-generated text.

The practical distinction that matters: features in this section will always produce the exact same output for the exact same underlying data (reproducible, auditable), while features in Section 3 may phrase the same underlying facts differently each time, since they involve a real language model call.

Which of the Section 3 live-AI features you can access depends on your subscription plan — for example, AI-generated fix explanations are part of Website Intelligence (available on every plan), while the AI Writer's content-generation tools are part of Content Intelligence, gated to the plans that include it. This is enforced the same way as every other plan-gated module in the product: a real, server-side check, not just a hidden UI element — see our Pricing page for which plan includes which module.

When a Section 3 feature runs, the relevant structured data already computed by our backend (and, for the AI Assistant, your chat messages) is sent to Anthropic's Claude API to generate the response. See our Privacy Policy, Sections 7 and 9, for what is sent and Anthropic's role as a sub-processor.

On AI training, stated plainly: Groager uses Anthropic's commercial API product (not the consumer Claude.ai chat product). By default, Anthropic does not use data submitted through its commercial API to train its models. Our codebase contains no configuration for, and we have not enabled, any Anthropic feedback-sharing or model-improvement opt-in program that would change this. Anthropic's own commercial API terms and Privacy Policy are the authoritative source for this, independent of anything stated here.

Live AI features (Section 3) depend on Anthropic's API being reachable and responsive. If that API is temporarily unavailable, rate-limited, or returns an error (including, in the past, exhausted API credits on our end), the affected feature is designed to fail gracefully — showing you an honest error or unavailable state — rather than silently returning a fabricated or default response. We do not guarantee uninterrupted availability of AI features, and Section 4's deterministic, rule-based features are unaffected by any AI-provider outage, since they don't depend on a live model call.

Groager currently uses Anthropic's Claude API for all Section 3 features. We reserve the right to change AI providers, use multiple providers, or change which specific model we use, at our discretion — for example, for cost, performance, capability, or availability reasons. If we change providers, we will update this policy's references accordingly; the Evidence Before Intelligence design principle (Section 2) applies regardless of which underlying model or provider powers a given feature.

There is currently no AI-specific request-rate limit layered on top of the general plan-gating described in Section 5 — we state this plainly rather than claim a limit that doesn't exist in the product today. Your practical usage is constrained by which AI features your plan includes, not by a separate per-minute or per-day cap on AI calls. We may introduce dedicated AI rate limits in the future (e.g. to prevent abuse); if we do, we will update this section to describe them accurately.

AI-generated output (Section 3) is intended to be reviewed by a human before you act on it, publish it, or send it externally — this applies especially to AI Writer content drafts, social post drafts, and any AI-generated recommendation with a real business, financial, or public-facing consequence. Groager does not auto-publish or auto-send AI-generated content anywhere on your behalf without your explicit action.

You are responsible for what you type into the AI Assistant or any other AI-powered input field. Do not submit information you are not authorized to share (e.g. another person's confidential information, or content that infringes someone else's rights), and do not use these features for any purpose that would violate our Terms of Service's Acceptable Use section.

Like all current large language models, Claude can produce output that sounds confident and plausible but is factually wrong — commonly called “hallucination.” Our Evidence Before Intelligence design (Section 2) is specifically intended to reduce this for the numbers and facts our backend controls (the AI is not permitted to invent a metric or data point that isn't already in the evidence we give it), but it cannot eliminate the general risk inherent to any AI-generated explanation, especially free-form text like AI Writer drafts or AI Assistant chat responses. Independently verify anything material before relying on it.

AI models, including Claude, are trained on large datasets that can reflect biases present in that data. AI-generated text (explanations, content drafts, recommendations) may reflect such biases despite Anthropic's own safety work. We do not independently audit Claude's outputs for bias before they reach you, and encourage you to review AI-generated content — particularly anything customer- or public-facing — with that in mind.

Beyond the Evidence Before Intelligence design constraint described in Section 2 (which limits what the AI is allowed to assert as fact), Groager relies on Anthropic's own model-level safety training and usage policies for the underlying AI behavior. We do not operate a separate content-moderation or safety-filtering layer of our own on top of Claude's output today. If you encounter AI output that seems unsafe, inappropriate, or in violation of Anthropic's usage policies, contact us at groagerai@gmail.com.

No AI feature of Groager — real (Section 3) or rule-based (Section 4) — guarantees any specific business outcome. This includes search engine rankings, organic or paid traffic, appearance or citation in AI-generated search overviews, leads, conversions, or revenue. AI-generated recommendations reflect our analysis of the evidence available at the time; whether implementing them produces a particular result depends on many factors outside the Service's control. Where we do not have sufficient evidence to responsibly estimate an outcome, the Service is designed to say so (“insufficient data”) rather than provide a speculative number — see also our Terms of Service, Section 11.

As between you and Groager, you own the AI-generated output you create using the Service (e.g. an AI Writer draft, a generated social post), subject to your compliance with these policies and your subscription. This does not grant you rights in Groager's underlying software, prompts, or models, nor any exclusive right against Anthropic or any other party.

AI-generated text is produced by a statistical model and may, on occasion, closely resemble existing third-party content it was influenced by during training, without our or your knowledge. We do not warrant that AI-generated output is original, non-infringing, or free of any third party's intellectual property rights. If you plan to publish AI-generated content publicly, you are responsible for reviewing it for originality and any copyright concerns before doing so, the same as you would for any other externally-sourced draft.

Nothing generated by any AI feature of Groager — real or rule-based — constitutes legal, financial, tax, medical, or other professional advice, and should not be relied upon as such. Independently verify anything material before acting on it.

As we add, change, or remove AI-powered features, we will update this page to keep the Section 3 / Section 4 breakdown accurate — reflecting what the product actually does, not a future roadmap.

Questions about this AI Usage & Disclaimer Policy can be sent to groagerai@gmail.com. We currently use this single address for all inquiries; a dedicated address for this policy is a planned future improvement once we have email hosting set up for our own domain, not yet in place today.