Andrej Karpathy, described in the source report as an OpenAI founding member, has suggested a practical way to improve the quality of large language model responses: ask the model to explain topics using a specific writing standard or output format. According to the report, examples include aerospace-style documentation, diagrams, HTML pages or custom explainer videos.

While this is not a formal product announcement or regulatory change, it is relevant for businesses experimenting with generative AI because it points to a simple operational lesson: output quality often depends heavily on how clearly companies specify structure, audience and format.

What happened

Search Engine Journal reported on comments from Andrej Karpathy about prompting LLMs to produce more useful answers by imitating structured communication styles. One example mentioned in the report is asking an LLM to write like an aircraft manual, which implies precise, step-by-step and unambiguous language.

The report also says users can request other output forms such as diagrams, HTML pages or custom explainer videos. These examples reflect a broader trend in generative AI use: moving from simple chat responses to more controlled, task-specific outputs that can fit directly into business processes.

Why it matters for European businesses

For SMEs and digital teams, the main takeaway is practical rather than theoretical. Many businesses using LLMs still rely on short, generic prompts and then find the results inconsistent. More structured instructions can improve reliability for tasks such as internal documentation, support content, training materials, product explanations and workflow automation.

This matters especially in business settings where vague wording creates operational risk. A technical support guide, onboarding checklist or compliance summary often needs consistency more than creativity. Asking an LLM to follow a formal style can help teams generate drafts that are easier to review, standardise and reuse.

There is also a digital operations angle. If an LLM can produce structured HTML or clearly segmented content, marketing and web teams may be able to accelerate landing page drafts, FAQ content, knowledge base articles or internal tools. That does not remove the need for human review, but it can reduce manual formatting work.

For European companies adopting AI, this fits into a broader shift from experimenting with chatbots to building repeatable AI-assisted workflows. Prompt design is becoming part of operational know-how, especially where businesses want outputs that are easier to validate and integrate into existing systems.

Who may be affected

• SMEs and founders looking for practical ways to get more dependable results from generative AI without investing in custom model training.
• Marketing teams producing structured content, product explainers, FAQs or web page drafts.
• IT and digital teams using LLMs for documentation, internal knowledge bases and workflow support.
• E-commerce businesses that need consistent product guidance, support content and customer education materials.
• Agencies and automation teams designing AI-assisted content and business processes for clients.

What companies should consider

• Define output standards clearly. Instead of asking for a general explanation, specify tone, structure, audience and format. For example, request step-by-step instructions, decision trees or page-ready HTML.
• Use AI for draft generation, not unchecked publication. Technical, legal or customer-facing materials should still be reviewed by a qualified human before use.
• Test repeatable prompt templates. Businesses that use LLMs regularly may benefit from creating standard prompts for support articles, campaign briefs, onboarding documents or product pages.
• Match style to risk level. Highly structured prompting is particularly useful where accuracy and consistency matter more than creative expression.
• Check AI outputs for compliance and factual accuracy. If content touches regulated claims, privacy, contracts or technical instructions, companies should not assume the model output is correct.

Overall, the reported comments do not represent a new platform release, but they do underline an important operational point for businesses: better instructions can lead to more useful AI outputs. For European companies trying to turn generative AI into a reliable business tool, prompt structure is becoming a practical capability rather than an experimental detail.