Skip to main content
Oracle is Athena’s accuracy workspace. It looks at how AI search engines (ChatGPT, Perplexity, Claude, Gemini, Google AI Overview, Copilot, Grok, and others) describe your brand, compares those statements against your verified facts, and surfaces every discrepancy it finds. From there, you can review each one, decide which version is correct, and even generate ready-to-publish content drafts that address the most important issues. You’ll find Oracle in the left sidebar.

​
What you can do here

Oracle is a single overview page. It shows:
  • Accuracy snapshot - discrepancy rate, open issues, and coverage trends.
  • Top high-impact discrepancies - the most important issues to review first.
  • Validation - external sources that contradict your verified facts after you run Verify facts.
  • Coverage - how complete your Pillars are relative to the topics AI is actually discussing.
A Run analysis button in the header lets you kick off two kinds of analyses on demand:
  • Check responses - scans AI responses already collected for your site and flags claims that don’t match your Pillars.
  • Verify facts - picks selected Pillars and looks for external sources (e.g. third-party websites, AI responses) that contradict them.
The Pillars and facts Oracle uses as its source of truth live in the Pillars tab of Knowledge Base. Oracle-generated remediation drafts are reviewed from Content > Content Hub as in-progress drafts.

​
How to get started

  1. Make sure the Pillars tab in Knowledge Base has at least the topics you care about. Athena uses your Pillars as the ground truth for every discrepancy check, so the quality of Oracle’s findings depends on what’s in there. If you haven’t set up Pillars yet, start with the Knowledge Base / Pillars article.
  2. Open Oracle in the sidebar.
  3. Click Run analysis > Check responses.
  4. Pick a date range (Last 7 / 14 / 30 days, or custom). Athena will show you how many AI responses fall in that window before you start.
  5. Click Start. The analysis typically takes 5-15 minutes. You can leave the page - a progress card stays anchored in the bottom-right corner and you’ll get notified when it finishes.
  6. When it completes, head back to Oracle to see your discrepancy rate, open count, and the top items to address.
You don’t need to configure anything else to get started - Oracle uses the AI responses Athena is already collecting for your site.

​
Reviewing discrepancies

When Athena finds a mismatch between an AI response and your Pillars, it shows up as a discrepancy. The Top high-impact discrepancies section on the Oracle page is the recommended place to start - it surfaces the most critical, highest-confidence findings first. Click any row (or See all for the full list) to open the Oracle Review Modal. For each discrepancy you’ll see:
  • Two claims side-by-side - what the AI said vs. what your Pillars say.
  • Why this was flagged - a short explanation of the conflict, with clickable references to the AI response and any supporting sources.
  • Research verdict (when available) - an extra AI pass that goes out to the web to determine which claim is correct, with cited sources.
You can then:
  • Approve one claim as the correct version (keyboard: 1 or 2 to pick, then Enter).
  • Mark both as true if the situation is genuinely nuanced (keyboard: B).
  • Skip to look at the next one (keyboard: >).
Approving updates the underlying Pillar claim so future analyses use the corrected fact. Notes are optional.

​
Verifying your facts against external sources

The Verify facts flow goes the other direction: instead of asking “what is the AI getting wrong,” it asks “what websites and AI responses are out there contradicting us?” This is useful for spotting outdated facts in Wikipedia, in competitor blogs, or in AI responses that haven’t picked up your latest information yet. To run it:
  1. Click Run analysis > Verify facts.
  2. Select the Pillars you want to check (you can pick a few, or select all).
  3. Click Start.
When it finishes, the Validation card on the Oracle page shows a ranked list of source domains by contradiction count. Click any row to see the specific contradictions for that domain and decide whether to update your Pillars or treat the external source as outdated. A pinned Knowledge Base row at the top represents internal inconsistencies - places where two of your own Pillars contradict each other. This is where to clean up before sharing your Pillars more broadly.

​
Drafting fixes

Oracle remediation drafts are reviewed from Content > Content Hub, not from a tab inside Oracle.
  • Oracle-created drafts appear in Content Hub as in-progress rows labeled Oracle.
  • Open a draft to preview the recommended content and see which discrepancies it addresses.
  • From the preview, you can Approve & publish to make the draft available for editing and publishing, or Dismiss if it’s not useful.
Approved drafts stay in Content > Content Hub, where you can edit, publish, or push them to your CMS through any connected integration (Webflow, WordPress, Shopify, etc.).

​
Reading the Oracle page

The Oracle page is your at-a-glance dashboard:
  • Discrepancy rate - share of analyzed AI responses with at least one claim that doesn’t match your Pillars. Lower is better. A small trend line shows how it’s moved across recent runs.
  • Open discrepancies - total unresolved discrepancies, with a sub-label showing how many are high-impact.
  • Validation card - appears once you’ve run a Verify facts analysis. Lists external source domains contradicting your facts.
  • Top high-impact discrepancies - the five most critical unresolved items, shortcut-clickable into the review modal.
  • Coverage - how complete your Pillars are relative to the topics AI is actually talking about. Shows total claims, claims AI has cited, uncovered claims (Pillar gaps), and trends over time. Two extra tabs under Coverage drill into specific gaps and unused content.

​
How Oracle and Pillars work together

Oracle is the audit layer; Pillars are the source of truth. Every discrepancy Oracle finds is a comparison between an AI response and the claims in your Pillars, so:
  • The more complete your Pillars, the more discrepancies Oracle can find.
  • Approving a side of a discrepancy updates the underlying Pillar claim, so your source of truth stays current.
  • Verify-facts runs use your Pillars as the input; the more authoritative they are, the more useful the validation report.
If you haven’t set up Pillars yet, do that first. The Knowledge Base / Pillars article covers everything you need.

​
Frequently asked questions

Do I need Pillars set up before running Oracle? You’ll get more out of Oracle when your Pillars are populated, but you can still run a Check responses analysis without them - it just won’t find as many discrepancies. We recommend running Discover pillars in the Pillars tab of Knowledge Base first. How often should I run an analysis? For most customers, a weekly Check responses run on a rolling 7-day window is a good cadence. Verify facts can be run less frequently - once a month is plenty unless you’re publishing rapidly. Why is my discrepancy rate going up? That usually means AI models have started repeating new claims that aren’t yet covered or corrected in your Pillars. Look at the top high-impact discrepancies - these are typically the source of the change. Approving the correct side of each will improve the rate on the next run. Can I undo a discrepancy resolution? Yes. Pillar claim changes are reversible from the Pillars tab in Knowledge Base. Open the relevant pillar, find the claim, and edit it back. What does “high-impact” mean? High-impact discrepancies combine three things: high severity (the claim is materially wrong, not just imprecise), high confidence (Athena is sure the claim is in the response), and high impact (the claim appears frequently or in influential responses). Start here. Where do approved drafts go? Approved drafts are saved into your Content > Content Hub as drafts. From there you can edit them and publish via any connected CMS integration (Webflow, WordPress, Shopify, etc.). Why don’t I see Fact Bank or Drafts in Oracle? Fact Bank now lives in the Pillars tab of Knowledge Base, and Oracle-generated drafts are reviewed from Content > Content Hub as in-progress drafts. Oracle itself shows the analysis overview and review workflows. Reach out to your account manager if you want to upgrade to Oracle. Can I cancel a running analysis? Yes. The progress card in the bottom-right of the page has a Cancel button. You can also dismiss the card to hide it - the run continues in the background and you’ll be notified when it finishes.