Content Strategy

What a Content Strategy is

A content strategy is the plan for what a business publishes, for whom, and how it will know whether the content worked. It has three parts that are easy to blur. The first is the choice of which questions the business will answer in public. The second is the way each page is built, so that the answer can be found and used. The third is the measure of success, which decides what the business looks at each month and what it does when the numbers move.

For a long time the third part was simple. A business tracked where its pages ranked and how much traffic they sent, and the plan was judged on those two lines. The site’s articles argue that this is no longer enough. Generated answers now sit between the question and the page, so the plan has to say how the business will be chosen as a source, not only how it will be found.

Why AI Answers change a Content Strategy

The ranking article says AI search visibility is where most content strategy goes vague. Its reason is that an AI Overview and a standalone assistant such as ChatGPT have never pulled from pages in the same way, and the gap between them has not stayed fixed. A plan written around one number called “AI visibility” misses that there are two systems behaving differently.

The AI Overviews article adds the underlying change. Ranking first once meant a click, because the searcher had to visit a page to get the answer. A generated summary removes that for many informational questions. The article says visibility has split into two questions: does the page still rank, and does it get cited as one of the sources behind the generated answer. A plan that optimises only for the first is aimed at a mechanic that now decides fewer outcomes.

This does not make the plan unnecessary. It changes what the plan is for. The pages still need to exist, be accurate and answer real questions. What moves is the effort spent on each one.

Where the effort goes

The AI Overviews article is specific about where the marginal effort should go. It names four places.

• Depth on narrow questions. Specific, long-tail questions earn citations that a broad chase of the biggest phrase will not. Under query fan-out, a page needs to be the clearest source for one of the sub-questions behind a search. The ranking article makes the same point with an example: a page that covers several angles of a topic in real depth can be pulled into more of the merged results than a page built around one phrase.
• Authorship and experience. Clear authorship and demonstrated experience decide whether a page is even in the running. The article puts it as a precondition: being selected as a trustworthy source comes before being cited. E-E-A-T is the name for those signals.
• Direct answers. A plain answer near the top of a section, a labelled list or set of steps and a claim that stands on its own are what a summarising model can lift cleanly.
• Branded search. The query type a generated summary is least likely to intercept, built over time and not in a single sprint.

None of these is a new discipline. They are the habits good web writing has always rewarded, applied with more care. The cost of a plan that ignores them is not a penalty. It is that the pages are never chosen.

What makes a page worth citing

The citation article describes a peer-reviewed study of techniques that raised how often content was cited in generated answers. The page leaves its figures to the article, but its pattern is worth stating. The techniques that worked added verifiable, attributable information: specific data, quotations from named sources and citations of credible outside sources. The ones that did not were keyword stuffing and simply adding more words. The article’s conclusion is that the signal is data density and source credibility, not length.

That matters for a plan because it changes what “more content” should mean. A business short of time does not need to publish more. It needs to make the pages it has more specific and better sourced. The statistics density page covers the first half of that. The article also says the technique that gave lower-ranked content the biggest gain was citing external sources, which is why it tells businesses to treat newer or lower-ranked pages as opportunities and not as lost causes.

Structure a Model can lift

A model assembling an answer is doing extraction, not composition. The ranking article says it reaches for whatever it can lift without reinterpreting a paragraph. A direct answer near the top, a labelled table in place of a wall of prose and a clearly formatted question and answer block each hand the model a claim it can use as written.

The AI Overviews article makes a point that keeps this practical. A page written to be extracted cleanly by a model and a page written to be scanned quickly by an impatient person want nearly the same thing. It also says freshness and specificity count: between two similar sources, a model leans towards the one with a checkable date and a specific figure over one making the same claim in vaguer terms. The structured content page covers how to lay a page out for this.

Branded Search and the click that survives

When a generic question is answered in full on the results page, the click that used to go to the top-ranked page often does not happen. The AI Overviews article says what does not disappear is a searcher’s tendency to look for a business by name once they already know it. A model has far less reason to intercept a query that already asks for one named source.

For a content plan, this makes brand recognition a more direct search asset than before. A business nobody has heard of has nothing to fall back on once the generic version of its category’s question is absorbed into a summary. A business people already know how to name is still found on purpose. Because it is built over time, it belongs in the plan as a standing aim and not as a campaign. The branded search page covers it.

What to track besides rank

The old dashboard of rank against traffic does not show whether a page is being used. The two articles name the same three habits.

1. Impressions against clicks. Search Console separates the two for AI Overview appearances. The shape to watch is impressions holding or rising while clicks on the same query fall. The ranking article says that is not a tracking fault. It is the question being answered on the results page.
2. Branded and non-branded queries as separate lines. Blended into one total they hide the shift, because a page can lose ground on generic questions while holding on searches for the business by name.
3. Reading the generated answer by hand. Check what an AI Overview or an assistant says for the handful of queries that matter most, and which sources it names. A page that used to rank first can find a competitor cited in its place.

The ranking article adds a fourth point. Citation frequency needs its own tracking and cannot be inferred from rank position, because being found on a results page, being cited in an Overview and being cited by a standalone assistant are separate outcomes that move independently.

Starting with an Audit, not a rewrite

The citation article’s advice on starting is direct. The gap most organisations need to close is not a rewrite of their whole content strategy. It is a targeted pass adding the elements the research found to matter. It says to start with an audit to find the pages already performing reasonably well, since that content is best placed to benefit first.

The AI Overviews article agrees and describes most of the work as that audit. It checks what a site already has against a different question from the one search optimisation used to ask. Not only “does this rank”, but “would a model reading across the top sources choose this page as one worth citing”. It also says that being in the running to be cited is a separate matter from what gets content cited once it is, and a business that gets the first wrong never reaches the second.

Neither article argues for dropping conventional search engine optimisation. The citation article says the techniques for being cited build on those fundamentals and do not replace them, so a business with sound, well-structured content is closer to benefiting than one starting from nothing. The generative engine optimisation page covers the techniques.

A worked example

An illustration, not a case study. A small training provider has a plan built around one broad phrase for its main service. It tracks the page’s rank and its visits, and both look steady, so the plan is judged to be working.

An audit asks a different question. Reading the generated answer for the queries that matter, the provider finds the broad phrase is answered on the results page and other businesses are named as sources. In Search Console, impressions on that page hold while clicks drift down. Branded searches, tracked as a separate line, are holding.

The provider does not rewrite the site. It picks the pages that already do reasonably well and adds a plain answer near the top of each section, a labelled table where a paragraph was carrying a comparison, a specific sourced figure where a claim was vague and a named author. It adds a page for each narrow question people actually ask around the main service. It keeps tracking branded and non-branded queries apart, and checks the generated answers by hand each month. The plan has changed in emphasis and measurement, and the content is mostly the same content.

Mix-ups worth avoiding

Four confusions come up often enough to name. The first is treating a drop in clicks as a drop in ranking, when the article says it can be the question being answered on the results page. The second is reading content strategy as a call to publish more, when the study behind the citation article found length alone was not the signal. The third is assuming a page that ranks first will be cited, when the two outcomes move independently and a page can be cited without ranking first. The fourth is treating AI visibility as one number, when the ranking article describes two systems that behave differently and change at different speeds.

How we use it

The citation article advises starting with an audit to find the pages already performing reasonably well, since that content is best placed to benefit first. The AI Overviews article describes most of the work as an audit of what a site already has, checked against the question of whether a model would choose the page as one worth citing. Branded search, built over time, is the query type it says a generated summary is least likely to intercept.

If you would like to talk through where your own site stands on both questions, whether it ranks and whether it is cited, a discovery call is a quick way to start.

Sources

• Why ranking on Google isn’t enough anymore: AI search visibility as two systems, query fan-out, depth across a topic, direct answers, labelled tables and question blocks, impressions against clicks, branded and non-branded queries tracked apart, reading generated answers by hand, and citation frequency tracked on its own.
• How AI Overviews are changing what counts as SEO: the old rank and traffic dashboard, ranking and citing as two questions, where the marginal effort goes, authority as a precondition, structure a model can lift, freshness and specificity, branded search, and the audit as most of the work.
• What gets content cited in AI answers: the peer-reviewed study of citation techniques, data density and source credibility over length, lower-ranked pages as opportunities, techniques that build on search fundamentals, and the targeted pass over existing pages that beats a rewrite.
• Search engine optimisation, Generative engine optimisation, Query fan-out, Branded search, Structured content, E-E-A-T, Statistics density, AI citation and Backlinks: the neighbouring terms this page links to.