# Response-text redistribution review

Reviewed 11 August 2026 · Decision for dataset version 1.0: **withhold full answer text pending written permission**

This is a publication-risk review, not legal advice.

## Decision

Do not publish the full ChatGPT, Gemini or Perplexity answer text collected through DataForSEO in the public JSONL yet.

Publish the 150-record metadata JSONL instead. It contains the full prompts, returned model labels, collection times, response lengths, SHA-256 hashes, token counts and complete citation arrays, while setting `answer_text_included` to `false`. This preserves auditability without assuming rights that the public contractual chain does not clearly grant.

## Why

1. [DataForSEO's public Terms of Service](https://dataforseo.com/terms-of-service), updated 12 June 2026, define “Content” broadly to include data and text made available through the service or developed via the API. The terms restrict some downstream SERP-data uses but do not contain an explicit public clause assigning ownership of, or granting a right to republish, complete LLM response text to DataForSEO customers.
2. DataForSEO also states that certain services are resold. The public terms do not explain whether every upstream output-ownership grant passes through to the DataForSEO customer.
3. The upstream terms are more permissive for their **direct** API customers: the [OpenAI Services Agreement](https://cdn.openai.com/osa/openai-services-agreement.pdf) allocates output ownership to the customer; [Google's Gemini API Additional Terms](https://ai.google.dev/gemini-api/terms) say Google does not claim ownership over generated content; and [Perplexity's API Terms](https://www.perplexity.ai/hub/legal/perplexity-api-terms-of-service) allocate output ownership to the customer, subject to applicable law.
4. Those upstream provisions do not resolve whether Gadex is the relevant direct “customer” when the request and result flow through DataForSEO, nor do they clear rights in third-party material that may appear in generated answers.

The uncertainty is contractual, not a finding that republication is prohibited. The prudent release state is “permission unclear,” not “permission denied.”

## Permission required to release full text

Obtain written confirmation from DataForSEO covering this exact use:

> May a DataForSEO customer publicly redistribute, as a downloadable research dataset, the complete generated answer text returned by the ChatGPT, Gemini and Perplexity LLM Responses API endpoints, provided that API credentials and reasoning fields are excluded and the engines, models, collection date and DataForSEO access method are attributed?

The confirmation should address:

- commercial publication on `gadex.ai`;
- downloadable JSONL redistribution rather than display only;
- all three LLM Responses endpoints used in the study;
- whether upstream provider attribution or notices are required;
- whether citation-grounded passages or other third-party content require additional treatment.

If permission is confirmed, regenerate the JSONL with only the user-visible `message` text. Never include reasoning items, credentials, internal request payloads or authentication data. Run a final third-party-content and personal-data review before publishing.

## Public wording

Recommended note beside the download:

> The answer-level file includes reproducibility metadata, integrity hashes and cited-source arrays. Full generated response text is withheld while Gadex confirms downstream redistribution rights for responses accessed through DataForSEO.