Generative AI Leader Practice Question: A media company wants to use a generative AI…

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Confidence indicators

  Why it's wrong here

  Confidence indicators quantify the model’s certainty in its output, but they do not retrieve or verify external source material; the scenario requires a retrieval-augmented generation (RAG) system that grounds each claim in a cited document. This option is tempting because confidence scores can suggest reliability, and in a scenario where the goal is to filter low-certainty responses rather than enforce factual attribution, they would be the correct choice.

• ✓

  Grounding

  Why this is correct

  Grounding connects the model to authoritative external data sources, so generated marketing copy cites verifiable original material rather than relying on parametric memory alone. This directly satisfies the stem's requirement for accurate source attributions, since citations must reference retrieved documents the model can actually point to.

• ✗

  Chain-of-thought reasoning

  Why it's wrong here

  Chain-of-thought reasoning shapes intermediate reasoning steps; it does not retrieve or bind source documents, so it cannot guarantee accurate citations. It is tempting because it improves answer quality, yet attribution requires grounding or retrieval-augmented generation against an authoritative corpus.

• ✗

  Safety filters

  Why it's wrong here

  Safety filters block harmful or policy-violating output; they neither retrieve source documents nor attach citations, so they cannot deliver the attributions required. They are tempting because they govern model output quality, but grounding or retrieval-augmented generation is what actually supplies verifiable source references.

Go deeper

Related to this question