We rank at position 22 for "seo for news publishers". That query put 430 impressions in front of us in the last 28 days. It produced zero clicks.
That is not a hypothetical. It is our own Search Console data for 23 August to 19 September 2026, and it is a useful failure, because the same pattern shows up on nearly every publisher site we audit. Ranking is not the problem. Ranking on the wrong kind of page is.
The 430 impressions were landing on /industries/news-publishers/, a service page. Someone typing "seo for news publishers" is not shopping for an agency yet. They want to know what news SEO involves. Google matched the query to the closest page we had, the page failed the intent, and nobody clicked. Across the whole publisher query cluster ("seo for publishers", "publisher seo", "increase discover clicks for publishers" and six more) that single page collected roughly 649 impressions and converted none of them.
This post is the page that should have existed.
Publishers are running three ranking systems, not one
Web search, Top Stories and Discover are separate products with separate eligibility. Treating them as one pipeline is the most expensive mistake in the category.
Web search rewards what you would expect: relevance, links, page experience. Top Stories adds a recency and authority filter that changes hour to hour and is heavily entity-driven. Discover has no query at all, which is the part that breaks most analytics thinking. There is nothing to rank for, so there is nothing to diagnose with a keyword tool.
Our own numbers show the split. "google discover" sits at position 10.2 with 13 impressions and one click. "google discover traffic drop" is at 13.8 with six impressions. Those are small volumes, but they behave completely differently from the web-search cluster around them: different pages win, different updates move them, and the traffic arrives in bursts rather than a steady line.
Google's own Discover documentation is explicit that there is no way to opt in beyond being indexed and meeting content policies. That is not a loophole to find. It is a statement that the lever is site-level quality, not page-level optimisation.
The audit finding that repeats on almost every site
When we audit a publisher, the failure is rarely a missing NewsArticle block. It is a conflict inside one.
The pattern looks like this. The visible article date says one thing. The datePublished property says another, usually because the CMS writes the record-creation timestamp rather than the publish timestamp. The author field contains a plain text string with no sameAs and no author page behind it. Each of those is individually survivable. Together they tell Google the publisher cannot be trusted to describe its own content accurately, and that judgement is site-wide, not page-specific.
We ran the same discipline on ourselves when we audited 25 SaaS platforms for llms.txt compliance. Only 12 of 18 sites that published the file followed the specified section structure, and just 2 of 18 emitted the discovery header. Publishing the artefact is the easy half. Publishing it consistently is what actually gets read by a machine.
The fix order we use, in this sequence, because each step makes the next one measurable:
- Reconcile every date field against the rendered page. No exceptions, no "close enough".
- Give every byline a real author entity page, and link the markup to it.
- Cut or consolidate thin archive pages — tag pages with four posts are dead weight on topical authority.
- Only then touch headlines and images.
Most publishers want to start at step four because it feels like SEO. It is the step with the smallest effect until the first three are true.
Impressions without clicks are a title problem, not a ranking problem
Our own data makes this uncomfortable to ignore. We hold position 9.3 for "aradhana restaurant" with 39 impressions and no clicks. Position 9.4 for "aradhana" with 51 impressions and no clicks. A top-ten position generating nothing means the snippet is losing to nine other snippets, or the intent was never ours to win.
For publishers the same arithmetic runs at a different scale. A site holding position 8 across 200 article queries with a 0.4% click-through rate is leaving five figures of monthly sessions on the table, and no amount of additional ranking work recovers it. The lever is the title and the description, and the test is whether a reader can tell what they will learn before they click.
We rewrote our own hero and internal linking on grow-nexus.com two days before writing this, specifically so that industry intent lands on an industry page one click from the homepage rather than three. The effect on this query cluster is not measurable yet. We will publish the delta when it is, including if it is nothing.
What we are not claiming
This is a 23-post site with 3,046 impressions over 28 days. That is a small sample, and small samples make weak inferences look strong. The structural failures described above come from client audits at larger scale, but the click and impression figures are ours and only ours, with all the limits that implies.
Nothing here recovers a Discover collapse in a fortnight. Discover moves with core updates, and the honest planning horizon is six to twelve weeks of consistent work before the signal responds. Anyone giving you a shorter number is describing a sales cycle, not a search one.
If you are running a publication and the pattern in this post looks familiar (strong positions, flat clicks, Discover traffic that arrives and leaves without explanation) the diagnostic work is the same regardless of who does it. Our news publisher practice starts with the markup reconciliation above, because it is the only step where the before-and-after is unambiguous.
