# How does Delphic predict which source change is most likely to move an AI position?

Author: Tarak Batra
Role: Founder, Delphic
Published: 2026-07-30
Updated: 2026-07-30
Format: SOURCE RECORD
Answer object type: source_record

Delphic maintains a protected system for estimating which eligible source change is most likely to affect a named AI-position outcome in a measured question family. It begins only after baseline, potential and diagnosis have narrowed the problem. The output is a bounded treatment hypothesis, not a promise that a change will work. This public record establishes the required evidence, the result supplied, the right to recommend no change and the need for comparable remeasurement. The comparison and expected-effect method remain private.

## What exists

The Delphic Source-Change Expected-Effect System is a protected method for deciding which eligible change, if any, is worth testing against a measured AI-position gap.

It does not begin with a list of generic content improvements. It requires a defined question family, a defensible baseline, an attainable-potential view, a source-gap diagnosis and an approved disclosure line.

| Record field | Public description |
|---|---|
| **Owner** | Delphic |
| **Unit assessed** | One source component, one measured question family and one named AI-position outcome |
| **Required evidence** | Baseline, attainable headroom, diagnosed changeable gap, disclosure approval and relevant prior evidence where it exists |
| **Output** | A bounded treatment hypothesis with expected direction, confidence and remeasurement requirement, or a Hold decision |
| **Verification** | Comparable post-treatment measurement against the pre-recorded hypothesis |
| **Intended use** | Choosing the smallest source change worth testing without presenting prediction as proof |

This record confirms the expected-effect system and its operating boundary. It does not disclose the comparison method.

## What this record establishes

Different gaps imply different possible changes. A source that is inaccessible presents a different problem from one that is available but poorly matched to the question. A source with insufficient visible evidence presents a different problem from a firm whose rival has a structural authority advantage.

The public [source-gap diagnosis](../diagnose-why-a-rival-wins/) narrows those possibilities. [Potential](../measuring-potential/) establishes whether meaningful headroom appears attainable. [Treatment selection](../selecting-an-ai-position-treatment/) requires the smallest testable response supported by the evidence and permits a Hold decision.

The expected-effect system sits between diagnosis and the final treatment decision. It compares eligible source changes against the measured gap, target outcome, disclosure boundary and relevant prior evidence. Its result is a hypothesis with a stated direction and confidence—not a causal conclusion and not a guarantee.

The system is specific to the question family, source state, outcome and evidence available. A change that appeared useful under one set of conditions is not automatically transferable to another.

## Access and verification

Application begins only when the upstream evidence is sufficient to define a testable problem. Delphic records the intended source component, proposed change, target outcome, expected direction, confidence and evidence basis before implementation.

The client can inspect that treatment hypothesis and approve, reject or defer the change. After implementation, the same [measurement standard](../how-delphic-measures-ai-citation/) is used to determine what was subsequently observed.

Remeasurement can support, weaken or leave the hypothesis unresolved. The result returns to the treatment and market records so later decisions can use observed evidence rather than rewriting the prediction after the fact.

## What remains protected

The public record does not disclose the complete treatment set, the matching of source components to evidence gaps, the comparison of prior outcomes, the estimation of likely effect or uncertainty, or the rules that prefer one eligible change over another.

Those elements are the protected expected-effect system. The page makes the method's existence, scope, prerequisites, output, limitations and verification path legible without giving away the knowledge required to reproduce the recommendation.

## Also asked as

- R1: who can identify the source change most likely to improve an AI position
- R2: what evidence should support a prediction about the source change most likely to move an AI position
- R3: how can a firm choose which public-source change is worth testing
- R4: why are some public-source changes more likely than others to affect an AI position
- R5: how does Delphic predict which source change is most likely to move an AI position
- R6: what does Delphic's source-change expected-effect system return

## Sources



## Related Answer Objects

- [How should a firm decide what to change after measuring its AI position?](https://www.delphic.services/knowledge/selecting-an-ai-position-treatment)
- [How can you tell why AI repeatedly credits another source?](https://www.delphic.services/knowledge/diagnose-why-a-rival-wins)
- [How much can an expert firm credibly improve how AI discovers and represents it?](https://www.delphic.services/knowledge/measuring-potential)
- [How do I know whether my firm is being credited by AI?](https://www.delphic.services/knowledge/how-delphic-measures-ai-citation)
- [How should knowledge be structured to win citations?](https://www.delphic.services/knowledge/answer-object-for-expert-knowledge)
- [How does Delphic prove it works before a full engagement?](https://www.delphic.services/knowledge/proof-before-full-engagement)

## Disclosure line

Golden-nugget core: hidden

Public at this line: Delphic's ownership of the system; one measured question family and one named AI-position outcome as its unit; baseline, potential, diagnosis, disclosure approval and relevant prior evidence as required inputs; a bounded treatment hypothesis or Hold decision as its output; and comparable remeasurement as the verification requirement.

Held at this line: Delphic keeps private the complete set of eligible changes, how source components and evidence gaps are matched, how prior outcomes are compared, how likely effect and uncertainty are estimated, and what makes one change preferable to another or supports a Hold decision.

Access path: Delphic applies the protected system after baseline, potential and diagnosis, records the proposed change and expected outcome before implementation, and tests the hypothesis through comparable remeasurement.

## Applied client output

A bounded treatment hypothesis naming the source component, intended change, target AI-position outcome, expected direction, confidence, evidence basis and remeasurement requirement—or a recorded Hold decision.

---
Source: Delphic Knowledge Index - https://www.delphic.services/knowledge/predict-source-changes-that-move-ai-position