Keyword
A keyword is the word or phrase someone types into a search box, or the concept they raise in a question to an AI assistant. It is the point of contact between what a person wants and what your content offers. In SEO the term is literal: a string you target. In AI visibility, the same underlying language matters, but engines like ChatGPT, Gemini, and AI Overviews respond to full questions and topics, not isolated strings, so a keyword functions more as a signpost to intent than a slot to fill.
Why keywords matter
Keywords tell you the actual words your audience uses, which is rarely the language you'd choose internally. Target the wrong terms and you either attract people who were never going to buy, or you attract no one. In classic search, keywords still decide which queries a page can rank for. In AI-assisted search that mechanism has weakened: Ahrefs found only 6-8% overlap between the URLs ChatGPT cites and Google's top 10 for the same query, and about 80% of ChatGPT's citations don't appear anywhere in Google's top 100 results. That means keyword targeting tuned purely for Google ranking will not reliably carry over to AI citation. Keywords remain useful anyway, because they still map the vocabulary and the gaps in what you cover — they're a planning tool, not a ranking guarantee.
How keywords work
A keyword's usefulness comes down to three things. Search volume: how often people type or ask it. Search intent: whether the person wants to learn something, compare options, or buy right away. Competition: how many other pages or sources are already trying to answer it. Short, broad terms like "hotel" carry volume but murky intent and brutal competition. Longer, specific phrases like "hotel that allows dogs at Lake Constance" — long-tail keywords — bring fewer searches but far clearer intent. The practical move is to assign a keyword, or a tight cluster of them, to a single page so that page answers one unambiguous question rather than several vague ones.
Common mistakes
The oldest mistake is stuffing: cramming a term into a page as many times as possible on the theory that repetition equals relevance. Search engines and language models both recognize the pattern and discount the content for reading unnaturally. A second mistake is keyword cannibalization, where several of your own pages chase the same term and split your relevance instead of reinforcing it. A third is optimizing for volume over intent — ranking for a popular term that has nothing to do with what your actual customers need. And a common structural error is treating a keyword as if it were the whole plan: a word is a signal, not an answer. What actually earns citation or ranking is the content built around it that resolves the underlying question completely.
Relation to AI recommendations
AI assistants don't search on isolated strings the way a traditional crawler-and-index engine does — they work from full natural-language questions, and increasingly from retrieval over content that already demonstrates authority on a topic. Keywords still matter here, but as a map of vocabulary and subtopics rather than a target to hit repeatedly. Google's own guidance is explicit that no special markup, schema, or AI-specific file is required to be surfaced in AI Overviews or AI Mode, and it warns against writing content "for AI" as a separate exercise from writing it for people. The more durable lever is being mentioned and discussed elsewhere: Ahrefs found that brand mention frequency across the web correlates with AI citation rate at roughly 0.664, about three times stronger than backlink count. So the practical shift is this — use keyword research to find the real questions your audience is asking, then answer them thoroughly enough, and get talked about widely enough, that an AI system has reason to point to you.
Example
Imagine a small bike repair shop. The owner notices customers rarely search "bike shop" — they search "get e-bike battery repaired." That's his keyword: concrete, with unmistakable intent. He builds a dedicated page answering exactly that question — cost, turnaround time, which brands he services. The page now ranks for people with precisely that problem, and it gives an AI assistant something specific and confident to point to when someone asks where to get an e-bike battery fixed. A broad keyword like "bike shop" would never have reached that exact group of buyers.
Common questions
Are keywords still important now that AI search exists?
Yes, though their job has changed. AI systems respond to whole questions rather than isolated terms, and citation doesn't track search rankings closely — one analysis found only 6-8% overlap between what ChatGPT cites and Google's top 10. Keywords are still valuable for mapping your audience's language and the topics you need to cover, just not as a direct lever on AI citation.
How many keywords should a single page target?
Center each page on one main topic with one or two core keywords and a handful of closely related variants. What matters more than the count is whether the page fully and naturally answers the underlying question, rather than repeating terms to pad relevance.