# How can you tell why AI repeatedly credits another source? Author: Tarak Batra Role: Founder, Delphic Published: 2026-07-18 Updated: 2026-07-29 Format: CAPABILITY Answer object type: capability “The rival was cited” describes the loss; it does not explain it. Delphic first establishes whether the pattern survives repeated measurement, then compares the role each source receives, the claims associated with it and the public evidence visibly available to support that role. The result is not always a content gap. The rival may have a changeable evidence advantage, an authority the firm cannot legitimately reproduce, an unstable lead caused by ordinary variation or an advantage whose cause remains unresolved. Diagnosis tells the firm which of those situations it is facing before it pays to intervene or discloses anything valuable. ## A rival citation is the wound, not the diagnosis The same visible loss can have very different meanings. A rival may fit the question more exactly. Its public evidence may be clearer or more current. It may hold a statutory, primary-data or institutional authority the firm cannot reproduce. Or the apparent lead may not survive another measured run. Delphic begins with the complete observed position, not citation count alone. It examines whether each source is surfaced, credited, recommended or selected; what role and attributes the answer gives it; and which public evidence is visibly connected to those claims. The comparison then meets the commercial reality: what the firm genuinely knows, what it can legitimately claim and where its disclosure boundary sits. A citation is useful evidence because it can show which public source grounded an answer and route a buyer to that source. It is not the verdict. A firm can be cited yet framed as secondary, or mentioned without being used as the basis of the answer. ## The comparison needs more than the rival's page A defensible diagnosis joins three views: - **the repeated answer field:** how the loss varies across natural questions, runs and engines; - **the public evidence field:** which sources are visible, what they establish and how the answer uses them; - **the firm's legitimate position:** the authority it actually holds and the knowledge it is prepared to make public. Without repeated evidence, variation can masquerade as a pattern. Without the public field, a page-level difference can be mistaken for the cause. Without the firm's private authority and disclosure line, the diagnosis can recommend a claim the firm cannot support or should not publish. Some engines do not expose a usable source list. Delphic still analyses presence, recommendation, comparison and answer framing, but source-level conclusions are marked unavailable or lower-confidence. The missing source record is not filled in by guesswork. ## Four findings, four different decisions | Finding | What it means for the client | |---|---| | **Changeable gap** | The firm holds, or can credibly build, public evidence that may address a material difference. | | **Structural advantage** | The rival's role rests on an identity, mandate or authority the firm cannot legitimately reproduce. | | **Run variation** | The apparent lead is not stable enough to justify intervention. | | **Unresolved cause** | The difference is real enough to investigate, but the evidence cannot yet distinguish the explanation without a test. | The client receives the measured loss, the material difference, the confidence attached to the diagnosis and the constraints that should not be treated. Where a change is plausible, the record supports the smallest credible test rather than a general instruction to publish more. The four findings are public because they change the commercial decision. Delphic keeps private the full diagnostic categories, how evidence is compared and how much each signal matters, how meaningful a difference must be, and how uncertainty about cause is handled. ## Diagnosis narrows the test Diagnosis does not guarantee that a treatment will move an engine, and it should not pretend an observable difference is a proven cause. Its job is to rule out bad work, isolate the most defensible explanation and define what a controlled intervention would need to test. Comparable remeasurement then shows what actually changed. The value of diagnosis is not certainty before evidence; it is a much smaller, more defensible decision than the rival citation first appeared to demand. ## Also asked as - R1: who can diagnose why AI repeatedly credits a rival for an important buyer question - R2: what evidence should a firm see before approving work to win back an AI answer - R3: how can a firm tell whether a rival's advantage can actually be changed - R4: what explains why AI repeatedly credits one source over another ## Sources ## Related Answer Objects - [How do I know whether my firm is being credited by AI?](https://www.delphic.services/knowledge/how-delphic-measures-ai-citation) - [How much can an expert firm credibly improve how AI discovers and represents it?](https://www.delphic.services/knowledge/measuring-potential) - [How should knowledge be structured to win citations?](https://www.delphic.services/knowledge/answer-object-for-expert-knowledge) - [How does Delphic predict which source change is most likely to move an AI position?](https://www.delphic.services/knowledge/predict-source-changes-that-move-ai-position) ## Disclosure line Golden-nugget core: full Public at this line: The need for repeated multi-signal comparison, the four broad diagnostic findings, the role of visible source evidence and underlying legitimate authority, the lower-confidence boundary when sources are unavailable and the client-facing diagnosis. Held at this line: Delphic keeps private the full set of diagnostic categories, the comparison method, the rules for deciding which differences matter and how confidently they explain the result, how signals are combined, how unclear cases are resolved and how a response is selected. Access path: Delphic applies the held diagnostic system to one measured question market and shows which difference is material, what is plausibly changeable and what still requires a test. ## Applied client output A question-specific diagnosis showing the repeated loss, the material difference, the confidence attached to it, what is changeable, what is structural and the next test the evidence supports. --- Source: Delphic Knowledge Index - https://www.delphic.services/knowledge/diagnose-why-a-rival-wins