# How Public Relations and Marketing Work Together in LLM Search Canonical page: https://sunlocke.com/articles/public-relations-and-marketing-in-llm-search Page type: article ## Summary Public relations and marketing can create different kinds of evidence around a business, but neither guarantees search ranking, language model retrieval, or a particular answer. As people ask language models what to trust, the public record around a business becomes more valuable and more difficult to interpret. Marketing creates an invitation, while public relations creates context around the claim. Neither controls how a person, search engine, or language model interprets the record. The public record now travels with the message Marketing speaks in the company's chosen voice. Public relations places that voice in a wider social record. A useful public record can give people and some search systems more material to inspect, compare, and cite. It cannot guarantee what any system will say. Why marketing and public relations are converging A clear service page can explain a claim. A case study can show the work. An independent reference can add context. These signals may help ## Article details As people ask language models what to trust, the public record around a business becomes more valuable and more difficult to interpret. ## Main point Marketing creates an invitation, while public relations creates context around the claim. Neither controls how a person, search engine, or language model interprets the record. Author: Anna Kyriazakou Section: Culture Post type: Framework Post Story format: standard ## Article content ## The question behind the claim Marketing usually begins with the message a business chooses to send. Public relations concerns the wider setting in which that message is received, discussed, questioned, or corrected. A person may compare a service page, a case study, an independent mention, a public review, a founder statement, and a correction before deciding whether contact feels reasonable. Each item carries a different kind of authority and a different risk of bias. Google Search Central says its guidance is about creating helpful, reliable, people first content and asks creators to consider original value, clear sourcing, and verifiable expertise. That is an attributed standard for content quality. It is not a promise about a result in search or in a language model. ## What is known Known for one product. OpenAI says ChatGPT Search can use web sources and provide links or citations. OpenAI also warns that search results and citations can be incomplete, outdated, or incorrect. This documents one product's behavior and does not establish how every language model works. Reasonable inference: a clear service page and an independent reference can give a person more material to compare than a promotional claim alone. The inference should be tested in the real customer question, not presented as a guaranteed visibility effect. ## What remains uncertain Editorial speculation, labeled as such: the public record may become a more important part of how businesses are encountered as people use assistants to form a first impression. It may also become harder to see which source shaped an answer. That uncertainty makes corrections, attribution, and restraint more important. The cultural question is not how to make a business appear everywhere. It is what kind of record the business is willing to leave when the message is read outside its intended setting. ## A practical audit Choose one public claim that a new customer could reasonably misunderstand. Gather the owned page, the supporting evidence, the independent references, and any correction. Do not treat the collection as equally authoritative. Label each item as verified fact, attributed statement, reasonable inference, or editorial speculation. Rewrite the claim so its scope, evidence, date, and limitation are visible. Remove language that implies guaranteed visibility or influence. Ask a person who was not involved in writing the claim what they believe after reading the record. Compare their interpretation with the intended one. ## The trust question Illustrative observation: a new customer is comparing a service page, a case study, a public review, a founder interview, and a correction. The page says what the business offers. The case study shows one reported experience. The review gives a person’s account. The interview gives an attributed perspective. The correction shows whether the business can acknowledge an error. The positive implication is a richer basis for judgment. The negative implication is that more material can create more contradiction. The responsible response is to clarify and correct, not to pretend every source carries the same weight. Source: Creating Helpful, Reliable, People First Content. Google Search Central This page supports Google's guidance about original value, clear sourcing, and content made to benefit people. It does not promise search visibility, language model retrieval, ranking, or agreement. Reference: https://developers.google.com/search/docs/fundamentals/creating-helpful-content Source: OpenAI Help Center. Searching the web with ChatGPT This page supports the current description of ChatGPT web search, including its use of web sources and citations and OpenAI's warning that results and citations can be incomplete, outdated, or incorrect. It does not establish the behavior of other language models or guarantee inclusion, ranking, retrieval, or a favorable answer. Reference: https://help.openai.com/en/articles/9237897 ## Structured sections The public record now travels with the message Marketing speaks in the company's chosen voice. Public relations places that voice in a wider social record. A useful public record can give people and some search systems more material to inspect, compare, and cite. It cannot guarantee what any system will say. Why marketing and public relations are converging A clear service page can explain a claim. A case study can show the work. An independent reference can add context. These signals may help a person evaluate a business, but they do not guarantee inclusion in a search result or language model answer. The evidence language models and people can find Google describes its generative search features as using publicly accessible and crawlable content. Google also states that meeting its requirements does not guarantee crawling, indexing, serving, or inclusion in an AI result. OpenAI describes ChatGPT search as a system that may search the web and provide links to relevant sources. That supports a careful statement about source discovery. It does not support a universal claim about how all language models form impressions. The strongest defensible argument is that responsible public evidence gives people and some search systems more material to inspect, compare, and cite. That is an interpretation and an operating recommendation, not a promise. Options breakdown Owned evidence: a clear service page, a documented process, a case study, or a public correction. The business controls the wording and must carry the responsibility for accuracy. Independent evidence: a review, interview, community reference, or third party account. The business does not control the wording, and the evidence may be incomplete or critical. Attributed interpretation: a statement from a named person or organization. Attribution helps the reader distinguish the speaker’s view from a verified fact. It does not make the statement true by itself. Step by step Choose one public claim that a new customer could reasonably misunderstand. Gather the owned page, the supporting evidence, the independent references, and any correction. Do not treat the collection as equally authoritative. Label each item as verified fact, attributed statement, reasonable inference, or editorial speculation. Rewrite the claim so its scope, evidence, date, and limitation are visible. Remove language that implies guaranteed visibility or influence. Ask a person who was not involved in writing the claim what they believe after reading the record. Compare their interpretation with the intended one. Tradeoffs More public documentation can help a careful reader investigate. It can also expose contradictions that the business would rather keep out of view. Independent evidence can add credibility through distance from the business. It can also be wrong, outdated, or based on one person’s experience. Optimizing a page for a known question can improve clarity. Chasing an imagined language model answer can flatten the writing and encourage claims no source supports. Common mistakes Treating a large technology company as automatic authority instead of naming what its source actually says. Promising that a public relations campaign will make a business visible in language model answers. Calling a company’s own description independent proof. Leaving a correction hard to find or silently changing a claim without explaining what changed. How to confirm Search the exact customer question and read the underlying sources, not only the summary or answer. Ask whether a reader can tell who made each claim, when it was made, and what evidence supports it. Repeat the audit later and in more than one search environment if the decision matters. Treat any result as a current observation, not a permanent guarantee. Example walkthrough Illustrative observation: a new customer is comparing a service page, a case study, a public review, a founder interview, and a correction. The page says what the business offers. The case study shows one reported experience. The review gives a person’s account. The interview gives an attributed perspective. The correction shows whether the business can acknowledge an error. The positive implication is a richer basis for judgment. The negative implication is that more material can create more contradiction. The responsible response is to clarify and correct, not to pretend every source carries the same weight. When to escalate Escalate claims about health, safety, finance, legal status, performance, clients, awards, or public allegations to the person qualified to verify them. Escalate a correction when the original claim could affect a person’s rights, money, safety, or reputation. Ask for specialist advice when a public relations plan depends on disclosure, defamation, privacy, intellectual property, or regulated advertising questions. What a smaller business can build The useful question is not whether a business can make a language model mention it. The useful question is whether the business has made its important claims clear, accessible, attributable, and testable. Start with one claim that a new customer could misunderstand. Link it to the company's own evidence. Then identify whether outside evidence confirms, complicates, or disputes it. If the record is incomplete, improve the record before increasing promotion. Next steps: quick wins this week Choose one public claim and link it to the evidence a new customer would need to inspect it. Add the date, scope, attribution, and limitation to any statement that could otherwise sound universal. Next steps: deeper work this month Build a small public evidence register with the claim, owner, source type, date, confidence, and correction path. Review whether public relations activity creates useful independent context or merely repeats the business’s preferred language. Create a record worth trusting The public record may travel with the message, but no business controls the route or the final interpretation. Make the record worth inspecting, then let the evidence do its quieter work. Sources for public relations and search Google Search Central. Guide to generative AI features on Google Search. https://developers.google.com/search/docs/fundamentals/ai-overviews Google Search Central. Creating helpful, reliable, people first content. https://developers.google.com/search/docs/fundamentals/creating-helpful-content OpenAI. Introducing ChatGPT search. https://openai.com/index/introducing-chatgpt-search/ Source boundary. These sources support public web access, helpful content, and web search with linked sources. They do not guarantee that a search system or language model will find, trust, or repeat a business claim. ## Next action Audit one public claim: /contact