Regularly, an SEO professional launches a shock statement against keywords, like ‘f*** keywords’. In an opinion piece originally published on his blog Chris Green SEO, Chris Green acknowledges the criticism is sometimes legitimate, but he questions what those who reject keywords propose, and what should replace them.
Chris Green admits the reflex has its reasons. He says he has worked on campaigns whose success hinged on the rise or fall of a few keywords. When rankings became the goal in themselves, it was easier to comment on two lost positions than to check if the main commercial need was being met. In the most extreme cases, decisions were made to promote keywords rather than serve the business objectives they were meant to achieve.
Which page answers which intent?
The author takes the example of a query like ‘hotels near Heathrow with a shuttle’, meaning hotels near Heathrow airport with a shuttle. Behind this search, a traveler wants to know if there is a shuttle to the terminal, if it runs at the time of their landing, and if they need to book a seat. The query also indicates the vocabulary to use to be recognized by the user.
If customers are looking for a shuttle, a page titled ‘ground transportation solutions’ asks them for an unnecessary effort. Chris Green notes he has often seen companies target the way they define their products or services, rather than the way people search for them. According to him, the initial query then becomes the symbol of a series of questions, which can be answered across several pages, sometimes called a hub-and-spoke model or content cluster.
In this context, keyword mapping, or a keyword plan, becomes truly useful well after the initial research. The author imagines that six months later, a location page is rewritten, shuttle schedules move into a general FAQ, and a new page on transfers is published. Each change may seem logical, but their sum can leave the site without a clear page to answer the question that was previously addressed.
What to track concretely?
Chris Green believes that topics and semantic groupings, which can be established using embeddings, i.e., mathematical representations of the meaning of a text, are useful. He nevertheless reminds us that we always have to choose what we type into a search engine or input into a query addressed to an AI. For him, tracking a group of queries or prompts remains necessary, but the aggregation must be based on precise choices.
He prefers to track a few representative searches for each page, observing which URLs appear and if they answer the question. These observations allow us to know if a page fulfills its role, if another page competes with it, and where content lags behind competitors. Chris Green acknowledges that keywords remain present in the data, even if we no longer measure the success of each one individually.
What happened between the ranking and the result?
For the author, these indicators shed light on what a site is visible for, but they don’t say what happens next. This, he says, is the most significant weakness. Ranking has long been one of the easiest things to observe, and the industry struggles to replace it with a more useful measure. Yet, even when a user sees a site for a query, the link between that moment, the click, and a significant commercial event can be almost completely broken.
Chris Green writes that, if he were to revisit campaigns he deemed too focused on isolated terms, he would probably keep some of the same keywords. He would use them to decide which searches matter most, which pages should answer them, and whether those pages are still fulfilling their mission. When judging the campaign, he would want to know what became of the customers and clicks that were being sought.
