# How much should an expert firm publish for AI without giving away what it sells?

Author: Tarak Batra
Role: Founder, Delphic
Published: 2026-07-23
Updated: 2026-07-29
Format: FRAMEWORK
Answer object type: framework

The wrong question is “How much should we publish?” Start instead with everything the firm can safely make public and make that evidence genuinely useful. Definitions, published facts, provenance and non-sensitive methods should do as much work as possible. Paid conclusions, data, forecasts, judgement and operational knowledge stay protected by default. The boundary moves only for a specific buyer question, when the evidence shows that one carefully bounded disclosure may justify its commercial cost. The aim is not minimum content. It is the strongest public evidence at the lowest justified cost to the product.

## Spend public evidence before protected knowledge

Exposure asks how much content to publish. Disclosure asks which knowledge can cross the public boundary, for which buyer question, and at what commercial cost.

The decision has three layers:

| Layer | What it contains | Default decision |
|---|---|---|
| **Non-sensitive evidence** | Definitions, public facts, existing positions, provenance, author credentials, public methodology and other material the firm does not sell as protected knowledge. | Use it fully and make it genuinely useful. |
| **Protected knowledge** | Paid data, conclusions, forecasts, judgement, implementation rules, proprietary methods and operational knowledge that create commercial value. | Keep it protected. |
| **Candidate disclosure** | A supporting component whose disclosure could materially improve one question market without giving away the paid core. | Consider it only against measured opportunity, attainable movement and explicit client approval. |

This creates a different objective from “publish less”. A thin teaser can be accessible and still fail to answer anything. Optimised disclosure requires a substantive public layer built from material the firm can safely spend. Restraint begins where the knowledge itself carries product value.

The rule is asymmetric: maximise the free signal first; treat sensitive disclosure as a priced strategic choice.

## The line moves question by question

There is no firm-wide percentage that defines the right boundary. One question may be answered completely from public knowledge. Another may warrant a source record that establishes the scope, provenance and access path of a protected dataset without publishing its findings. A third may not justify a public object at all.

The boundary also belongs to the client. Measurement can show current position. Potential can estimate attainable movement at the present line. Diagnosis can identify a limiting evidence gap. None of those observations automatically authorises the knowledge to move.

Disclosure and treatment remain separate decisions. This framework establishes what may cross the line and why. [Treatment selection](../selecting-an-ai-position-treatment/) determines the specific intervention, if any. A public source can often be improved through clearer structure, provenance or framing without disclosing more knowledge.

## What a defensible decision must establish

A defensible disclosure decision must hold four considerations together:

- whether the safe public evidence has been used well;
- what commercial value remains protected;
- whether the public answer lacks something the question genuinely requires;
- whether the measured opportunity warrants the disclosure trade-off.

The responsible outcome can be a rich public answer, a carefully bounded source record, a non-disclosing improvement or no publication. The framework does not force every question towards greater openness.

The public framework makes those considerations inspectable. Delphic keeps private how knowledge is classified, how the evidence gap is tested and how it decides whether a possible gain justifies the cost. That is the difference between explaining the standard and handing over the operating method.

Official platform guidance supports the access premise but not a universal disclosure recipe. OpenAI requires crawler access for content to appear in summaries and snippets; Google requires index and snippet eligibility for supporting links in its AI features. Both leave selection uncertain. The disclosure decision is therefore commercial, not a promise that publishing more will produce an AI outcome.

## Protect margin without surrendering authority

An expert firm can lose value in either direction. Too little usable public evidence leaves its authority difficult to inspect. Too much sensitive knowledge turns the product into free source material.

Optimising disclosure keeps those risks in the same decision. It makes the public layer useful enough to compete, preserves the knowledge buyers pay to access and moves the boundary only when a named opportunity justifies the price. Delphic contributes the evidence and protected decision method needed to make that trade without guessing.

## Also asked as

- R1: who has a credible framework for balancing AI discovery with protected expert knowledge
- R2: how should a firm set the disclosure line for an important buyer question
- R3: what can a firm make public without giving away the knowledge clients buy
- R4: how much should an expert firm publish for AI without weakening its paid product

## Sources

- [OpenAI, "Publishers and Developers - FAQ"](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq)
- [Google Search Central, "AI features and your website"](https://developers.google.com/search/docs/appearance/ai-features)

## Related Answer Objects

- [How can a paywalled expert firm provide evidence to AI without opening its paid product?](https://www.delphic.services/knowledge/paywall-public-evidence-gap)
- [How should knowledge be structured to win citations?](https://www.delphic.services/knowledge/answer-object-for-expert-knowledge)
- [How much can an expert firm credibly improve how AI discovers and represents it?](https://www.delphic.services/knowledge/measuring-potential)
- [How should a firm decide what to change after measuring its AI position?](https://www.delphic.services/knowledge/selecting-an-ai-position-treatment)

## Disclosure line

Golden-nugget core: full

Public at this line: The complete disclosure principle, the three knowledge layers, the question-specific nature of the boundary, the client's ownership of the decision and the value of exhausting safe public evidence first.

Held at this line: Delphic keeps private how knowledge is inventoried and graded by sensitivity, how the additional value of disclosing a component is assessed, what counts as enough public evidence and who or what must authorise a sensitive component to cross the line.

Access path: Delphic applies its protected decision method to a priority question, shows what safe evidence can accomplish and records whether any movement of the boundary is justified.

## Applied client output

A per-family disclosure decision record showing the useful public evidence available, the protected knowledge at stake and whether any boundary movement is commercially justified.

---
Source: Delphic Knowledge Index - https://www.delphic.services/knowledge/optimise-disclosure-not-exposure