For many years, writing for the web was a soulless train in pleasing bots. We stuffed articles with key phrases, constructed pages for crawlers, and centered on search engine rating above all else. It was a recreation that wasn’t ultimate for anybody.
However that recreation is altering, quick. Now, individuals are more and more turning to ChatGPT and different massive language fashions (LLMs) for data. They’re not searching for an inventory of hyperlinks. They’re asking questions and anticipating a direct, clever reply. It is a profound shift: As an alternative of looking, customers are asking, they usually count on considerate, nuanced, and context-aware responses.
Meaning the previous playbook — constructed for rankings, not reasoning — is giving method to a brand new technique constructed on authority. Now, writing to be cited by AI means crafting skilled content material that will get shared by machines in conversations with their customers.
This considerations everybody, not simply your content material groups. This adjustments the way in which your organization communicates and, extra importantly, is found and understood.
How do you keep seen when your clients are skipping the search bar? Begin by writing not only for individuals or algorithms, but additionally for AI reasoning. Salesforce has been evolving its technique, and we’ll share six greatest practices in a bit.
The state of search
Google nonetheless dominates, driving greater than 90% of search site visitors, however AI use is rising quick. In response to a December 2024 Bain & Firm survey, 80% of customers resolve 40% of their on-line searches with out clicking a hyperlink. AI-generated summaries have already contributed to an estimated 15% drop in international search site visitors.
This shift is going on as a result of AI delivers direct solutions, not simply hyperlinks. The purpose for manufacturers is now not to win a click on, however to be the supply AI trusts sufficient to cite.
That’s why each firm that publishes content material must increase past key phrases to an LLM-first strategy.
“There’s one million issues you are able to do now to tweak a web page and have it present up in rankings and AI overviews,” mentioned Graham Hardt, senior director of search engine optimisation at Salesforce. “Persons are having conversations with LLMs, and that adjustments the way you map to the infinite variety of prompts that somebody would possibly ask.”
How LLMs discover and floor content material
LLMs excel at understanding detailed or particular questions, these long-tail queries that search engines like google and yahoo typically battle with. For instance, ”What’s one of the best ways to take away crimson wine stains from a silk shirt with out damaging the material, and what merchandise ought to I exploit?”
AI in search engine optimisation: A Full Information
The wedding of AI and search engine optimisation is a brand new frontier in digital advertising and marketing. On this information, you’ll learn the way AI improves key phrase search, how AI helps with optimization, and rather more.
LLMs can interpret these queries by analyzing the that means behind the phrases, and matching related data even when precise key phrase matches are uncommon. This implies they will join the dots throughout totally different items of content material and generate helpful, tailor-made responses.
Typically, LLMs use Retrieval Augmented Era (RAG) to tug in further data from verified sources. They don’t dump all the information on you, however select probably the most helpful bits to offer you correct and full solutions.
That’s why content material should be clear, well-structured, and so deep on particular matters. When your pages are organized logically and use constant patterns, LLMs can interpret them extra precisely, and floor them when individuals ask much less widespread, high-intent questions.
What’s the distinction between writing for search engine optimisation and writing for LLMs?
Writing for LLMs versus conventional search is totally different in three key methods.
1. Skimming vs. construction
Conventional search engine optimisation content material is written for people to skim, with quick paragraphs and bullet factors. When writing for LLMs, your content material wants stronger construction so algorithms can parse it. Use subheadings framed as actual questions that readers would possibly ask, just like the one above. Break matters into clear, digestible sections so LLMs can determine particular person information and information factors.
2. Reasoning vs. rating
The previous playbook was about getting your content material to rank on web page one in all search outcomes. The content material’s objective was to get readers to click on to your web site. In case your web page had a great title and was in a high place, you earned a click on.
To succeed in the present day, content material must be efficient for each conventional search and AI. You continue to have to rank excessive for conventional blue hyperlinks, however your content material additionally should be a supply for AI Overviews and LLMs like ChatGPT and Gemini. This implies your content material should be structured for reasoning, not simply rating. The knowledge must be so clear, factual, and well-organized that an AI can simply perceive, extract, and rephrase it into a great response.
“AI Overviews are, in some ways, an extension of your web site,” mentioned Hardt. “That’s how that you must give it some thought, and craft your content material accordingly.”
3. Key phrases vs. authority
“In the present day, efficient search engine optimisation is about offering probably the most authoritative, reliable, and useful content material for a given subject,” mentioned Hardt. “It goes far past merely utilizing key phrases. Success will depend on a holistic strategy, together with a powerful data structure and content material construction, demonstrating experience and authority (E-E-A-T), and constructing a precious back-link profile.”
The purpose, he says, isn’t to jam key phrases into content material, however to create a high-quality useful resource that earns its place on the high of an inventory of sources by offering a superior person expertise and actual worth.
In case you’ve achieved all this, you’ve additionally positioned your content material to be a trusted supply for AI search.
AI Overviews and LLMs prioritize authority and belief. They search probably the most credible, complete, and up-to-date sources to kind their reply. So, your purpose is to develop into a subject skilled, creating content material that AI can cite because the educated supply.
6 tips about writing for AI
With AI Overviews and LLMs reshaping how individuals discover data, writing completely for search rankings now not works. To remain seen and related, your content material should be price quoting instantly.
This shift calls for a companywide rethink of how you propose, create, and construction content material. Right here’s how we’re approaching it at Salesforce.
1. Write clearly and succinctly
Use easy, conversational language and clear, concise sentences. Embody “final up to date” dates to inform the LLM how just lately your content material was refreshed.
2. Assist algorithms belief you
LLMs reward content material that’s reliable and reveals experience. Hyperlink to respected sources, and embody creator bios demonstrating the author’s credentials. Add quotes, and refresh your pages each six months. Keep in mind, stale content material will get ignored.
3. Make content material simple to learn and perceive
Construction is essential. Use headings and subheadings, and a logical circulation. Body subheadings as linear questions readers (and AIs) would possibly ask, like “What’s an AI agent?” adopted by “How do brokers work?”
To make your content material much more efficient, add schema markup within the code in your web page. This language acts as a translator, explicitly telling search engines like google and yahoo and AI the that means of your content material, which makes you a extra doubtless supply for his or her solutions.
4. Use TL/DRs
LLMs love TL/DRs (too lengthy, didn’t learn). Start your content material with a brief, bulleted record of key takeaways that match the person’s intent. The TL/DR provides context, not simply navigation, serving to an LLM see your web page as related.
5. Use FAQs wherever attainable
An FAQ helps customers and LLMs shortly discover solutions to widespread questions, and boosts your likelihood of being featured in an AI Overview or a Google snippet. Use pure language that matches how customers ask questions. For concepts, see Google’s “Folks Additionally Ask” part or Reply the Public.
Salesforce is already taking steps with FAQs, and plans so as to add FAQs on all authoritative, product and pricing net pages by 12 months’s finish.
6. Assist, don’t promote
When individuals ask an LLM a query, they’re searching for useful solutions, not a gross sales pitch. Focus your content material on figuring out and fixing an issue.
Hardt additionally advises firms to analysis long-tail, pure language questions that individuals are asking.
“Lots of people are asking longer-tail questions like ‘What’s the greatest CRM software program in 2025 for small companies?’ A lot of these queries could have much less quantity, they usually could not get requested fairly often, however they display particular intent and allow you to know what individuals need,” he mentioned.
Authority and readability rule
Finally, this shift isn’t about know-how; it’s about belief and usefulness. The previous search engine optimisation recreation was a collection of tactical strikes to earn a click on. The brand new recreation is a mandate for authority, authenticity, and readability. It’s a return to the core ideas of fine communication, however with a brand new viewers: AI fashions that act because the gateway to a world of data.
What’s your agentic AI technique?
Our playbook is your free information to changing into an agentic enterprise. Study use circumstances, deployment, and AI abilities, and obtain interactive worksheets in your workforce.
