How to Prioritize Content Ideas That Drive Traffic and Results

The hard part of content planning is not finding ideas. It is deciding which ideas deserve a page, which pages already on the site deserve work first, and which topics should stay off the calendar.

Start with the site before the keyword list. A page already collecting impressions for the right queries is often a better opportunity than a new topic, because the demand is proven and the work is smaller. A strong idea with no firsthand knowledge behind it, and no clear route to a reader, may not be worth publishing at all.

This is the process we use. It ends in one of five decisions: publish something new, improve a page you already have, combine overlapping pages, retire a page, or send the idea back for research.

Start With the Pages You Already Have

Search Console holds the evidence for most of these decisions, and it is evidence about your site rather than an estimate about your market. Two questions are worth separating, because they are different problems with different fixes.

The first is about pages Google shows to people who do not click. The second is about pages Google has not indexed at all. Treating both as one SEO problem is how a team spends a month rewriting titles on pages that were never in the index.

Reading Impressions Against Clicks

Open Performance, then Search results, and set the range to at least three months. Six is better on a low-traffic or seasonal site. Switch to the Pages tab, turn on all four metrics, and sort by impressions.

Four combinations matter, and where a page falls tells you what kind of work it needs.

What impressions and clicks tell you together

Impressions

Many impressions · Few clicks

Strengthen this first

Google is showing the page and people are choosing something else. The demand is already proven, which usually makes this the cheapest work on the list.

Many impressions · Many clicks

Study the pattern

This is working. Read its query list for subtopics you have not covered, and look at what the page does that your weaker pages do not.

Few impressions · Few clicks

Investigate before deciding

Low impressions has several possible causes. Check whether the page is indexed, whether anyone searches for this, and whether another page is competing with it.

Few impressions · Many clicks

Leave it alone, and protect it

A narrow query with a high click rate. Rarely worth expanding, usually worth protecting when you reorganize.

Clicks

One caution about the position column, because it gets read wrongly more often than any other number here. Google defines average position as the topmost position your result reached, averaged across all queries the page appeared for. A page averaging position 12 may sit at 4 for one query and 40 for thirty others. It is a rough indicator of placement rather than evidence that a page sits on page two.

So do not conclude anything from the page row alone. Click into the page, switch to Queries, and read what Google is matching it against.

What the Query List Usually Shows

Three patterns come up repeatedly, and each points somewhere different.

The page ranks but answers late. The query asks something specific and the page takes four paragraphs to get there. Rewrite the opening so the answer arrives in the first hundred words.

The impressions are spread across unrelated queries. The page is trying to be about several things. Narrow it, or split off the part with real demand.

One query pulls in several of your pages. Sometimes that is fine and Google is choosing sensibly. Sometimes two pages answer the same need and neither is complete. That points at combining them rather than rewriting either one.

Read the Live Results Page

Search Console reports the outcome without showing you the page a searcher looked at. For any query worth acting on, run it yourself in a clean browser session and read the whole page before deciding anything.

Three things are worth writing down.

What format Google is rewarding. If the first several results are comparisons and the page you are protecting is a service page, the mismatch is the finding. Formats shift by query and they shift over time, so this is worth rechecking rather than assuming from what ranked last year.

Who you are competing with. It is frequently not the businesses you think of as competitors. A local service query can return directories, a marketplace, a national brand and one competitor. That changes what a realistic outcome looks like, and sometimes it is the reason to leave a query alone.

What sits above the first result. Ads, a map pack, shopping results, a video carousel, a featured snippet or an AI answer all take space and attention before anyone reaches an organic listing. A page ranking third on a query where an AI summary answers the question completely is in a different position than the same rank on a query that needs a decision made.

That last one connects directly to the click resilience question below. If everything a reader needs is already on the results page, the work is having something that page cannot supply.

Mining a Page That Already Works

The pages with clicks are the ones most teams stop looking at, and they are the best source of ideas on the site. Andy Crestodina’s content audit process starts from the same place, on the observation that “a small handful of articles pull in a big percentage of traffic and conversions.” Finding those pages and asking what they have in common is more productive than opening an empty calendar.

Read the query list on a page that performs. You are looking for two gaps. The first is subtopics the page raises but never covers in depth, usually visible as queries drawing impressions to a page that was never about them. The second is a missing parent or a missing child.

A cycling retailer might have four good posts on gravel tyre pressure, tubeless setup, sealant and rim compatibility, and nothing answering “how to set up a gravel bike for a first long ride.” The four detailed pages exist because customers asked those questions. The overview never got written because nobody asks it at the counter. It is the page a beginner searches for first.

The reverse happens too. A site with one broad guide to trail running shoes, and nothing on shoes for wide feet, wet-weather grip, or a first ultramarathon, is missing the pages where the decision gets made.

When Google Has Not Indexed the Page

The Pages report under Indexing answers a different question, and two of its statuses get read as the same thing. They are not.

Discovered compared with crawled, and what each one means

Aspect	Discovered, currently not indexed	Crawled, currently not indexed
What Google says	The page was found but not crawled yet. Google says this typically happens when crawling the URL was expected to overload the site, so the crawl was rescheduled.	The page was crawled and not indexed. Google says it may or may not be indexed later, and that there is no need to resubmit the URL.
What to check first	Whether the page has internal links from pages already indexed, whether it is in the sitemap, and whether the site is slow or rate-limiting crawlers.	Whether the page is meaningfully different from pages already indexed, whether the canonical is right, and whether the main content survives rendering.
What it is not	A verdict on the content. Google has not read it yet.	A permanent decision, and not something another indexing request will fix on its own.

The distinction decides where you spend the afternoon. Discovered points at links, sitemaps and server capacity. Google’s guidance on links is direct: every page you care about should have a link from at least one other page on your site. A page reachable only from a sitemap is one you have asked Google to find without giving it a reason to.

Crawled and not indexed is the harder one, and it is where the content question lives.

The Pages That Never Get Indexed

Google fetched these and did not include them. That can change, and a genuinely new page often needs nothing but time. But when it affects a whole template at once, the pattern is usually that the pages are not distinct from each other or from what is already indexed.

The version we see most often on client sites is a set of pages that are all true, all competent, and all interchangeable with what is already in the results. Every therapy practice site has a page explaining what EMDR is. Most of them say roughly the same thing, because the subject is well documented and each page was written from the same public sources.

Google has not published a rule about this, so treat what follows as our reading rather than a documented mechanism. There is a Google patent, Contextual estimation of link information gain, describing how a document might be scored for information it adds beyond documents a user has already seen. A patent is not published ranking guidance and should not be used as one. It is a useful way to frame the problem: if a page adds nothing to what a reader has already been shown, it is hard to argue it needs to exist.

The practical test is not about search at all. For each page in that group, ask what a reader gets here that they would not get from the results already ranking. A treatment page can answer what happens in a first session at this practice, who it is not appropriate for, what it costs, and how long people usually stay. That is firsthand, specific to the business, and no other site can write it.

If the honest answer is that the page repeats what is already available, another crawl request will not solve it. Combining the page into a stronger one, or keeping it out of the sitemap and linking it where it is genuinely useful, gets you further.

Four Questions Before Any Scoring

Now the ideas. Run each candidate through four questions in order. Most get resolved before any scoring starts, which is the point.

1. 1

   Does this support a real decision?

   Someone choosing, comparing, using or renewing something the business sells. If not, take it off the plan rather than scheduling it.

2. 2

   Can we name the evidence that makes this page ours?

   A firsthand implementation, original data, screenshots, a subject-matter interview, or the questions customers keep asking. If none of those exist yet, the idea is not rejected. It is marked research needed.

3. 3

   Do we already have a page that partly serves this?

   If yes, the decision is refresh, combine or redirect before it is create. A second URL competing with your own page helps nobody.

4. 4

   Can we say how a reader will find and use it?

   Name the internal links it will get. Name the person who keeps it accurate after publication. Say what you expect to change. An idea missing all three is not ready.

The second question is the one that changes a content plan most. An idea with no evidence behind it is usually unresearched rather than weak, and the difference matters: “research needed” sends someone to find a source, while “rejected” throws away a topic that might have been the best on the list. Google’s own guidance asks whether a page offers original information, reporting, research or analysis, which is the same question asked from the reader’s side.

The third question is the one most planning processes skip, and it is where the cheapest wins usually are. Refreshing a page that already has impressions is smaller work than a new post and starts from proven demand.

Two things here get overstated, so worth being accurate about. An existing URL keeps its links and its history, which is a fact. It does not follow that the page will be easier to rank, and a page whose intent no longer matches the query may be better replaced than repaired.

Ask What Survives an AI Answer

One dimension belongs in this decision that was not part of it a few years ago. When a search result or an AI summary answers the question on the page, what is left that still requires a visit?

Aleyda Solis calls this click resilience, and her content prioritization framework for AI search puts it alongside citation potential, brand mention potential, business value, proprietary advantage and expected effort. Her framing of the test is the useful part:

The key question is what remains unresolved after the AI answer and whether your content is the best resource to satisfy that remaining need.

Her framework also separates click value, citation value and business value rather than collapsing them into one number, which matters because a page can serve one well and the others barely at all. A piece of original research may be cited widely and send few visits. A pricing page may send few people anywhere and still be one of the most valuable pages on the site.

In practice, the content that holds up tends to include current prices, availability or policies, a task the reader has to complete, a comparison with the criteria disclosed, firsthand testing, or guidance that depends on the reader’s own situation. The content that does not tends to be a definition, a basic list, or a general guide assembled from what was already published elsewhere. That maps closely onto the pages sitting in crawled and not indexed.

Score Only What Survived

Only now is a score useful, and only as a tie-breaker between candidates that already passed. Rate each input from 0 to 3:

Input	The question it answers
Business fit	Does this help someone choose, use, compare or renew what we sell?
Distinct evidence	Do we have firsthand material nobody else can publish?
Click resilience	After a summary answers the question, what still requires a visit?
Existing signal	Does a page already get relevant impressions, links or sales questions?
Distribution path	Where will a reader meet this, other than search?
Creation effort	Research, interviews, writing, design, review
Maintenance burden	How fast does it go stale, and who reviews it?
Cannibalization risk	Would another URL make the site less clear than it is now?

Add the first five, divide by the last three, rank what is left.

The number is a tie-breaker, not a decision. An idea with no business fit and no original evidence does not become worth publishing because it is cheap to produce. That is the failure this ordering exists to prevent: a short AI-assisted summary costs almost nothing and can still add nothing, and a scoring model that rewards low effort will keep recommending it.

We are not claiming the formula is validated. Scoring models of this shape come from product planning, where they rank features, and no published research says the weights carry over to content. It makes a judgment inspectable without turning it into a measurement.

Where AI Fits, and Where It Does Not

Generating ideas is now close to free. A model will produce forty plausible titles in a minute, and none of that is evidence that a reader needs any of them.

Treat an AI list as prompts for research rather than a plan. The four questions still apply, and the second one is the filter: if the only source for a piece is what a model already knows, then by definition it is material a reader can get elsewhere.

Where it earns its place is on bounded work inside a decision already made. Grouping a query export into themes. Drafting the counter-argument to a section so you can see what you left out. Turning a process you know well into a first outline. Suggesting where a comparison or a diagram would carry an idea better than another three paragraphs.

Write Down Why, and Check in 90 Days

Record the decision when you make it, with the evidence you had at the time. A short log is enough:

Candidate	What we had	Decision
Old service guide	Impressions on the right queries, an existing page with few clicks, and the same question coming up on sales calls	Refresh, recheck in 90 days
Broad explainer	No firsthand source and no original data	Research needed, revisit when a customer asks
Platform comparison	Builds on both platforms, with screenshots and real timings	Publish, recheck in 90 days

Record who owns accuracy after publication, too. A page nobody owns is a page that goes stale on the schedule of whoever last noticed.

Then look again after about 90 days and compare what you expected with what happened. The next decision then rests on a record rather than on memory.

Set the recheck date when you make the decision. A review that is not scheduled happens the next time something looks wrong, which is always later than it needed to be.

What to Do This Week

The next content decision is as often what to improve, what to combine, or what to stop publishing as it is what to write.

Open Search Console and sort your pages by impressions. Find the ones being shown and not clicked, because that is the shortest distance between work and result. Then look at what is not indexed, and be honest about which of those pages add anything a reader cannot already get.

Start with the pages and the questions the business already has, then pick the smallest piece of work that moves the reader or the business forward.