Google Discover Search Console Report: Data, API and Traffic

What the Discover report in Search Console shows

The Discover performance report in Google Search Console shows how often your pages appeared in Google Discover (impressions), how often people tapped them (clicks) and the click-through rate, for up to 16 months, with days in Pacific Time. It has no queries and no average position, because Discover recommends content without anyone searching. You’ll find it under Performance → Discover, but only once your property passes a minimum number of Discover impressions. The same data is available through the Search Console API with the search type set to discover.

Discover is the feed of articles and videos in the Google app and Chrome, picked from each person’s interests rather than from a query. Google says it regularly shows sports, health, entertainment and lifestyle content, so for those publishers it can be a large, unpredictable share of Google traffic.

We make GSC MCP Server, a read-only connector that lets AI assistants query Search Console, so some steps below use our tools. Every check also works in the Search Console interface. Google facts come from its own documentation, checked on 30 October 2026.

What’s in the report

• Metrics: clicks, impressions and average CTR. An impression counts when a card scrolls into view, once per result per session; sharing a card is not a click.
• Attribution: every click and impression goes to the canonical URL, and appears only in the property that contains it.
• Coverage: the Google app and, since February 2021, Chrome.
• Freshness: data arrives within 2–3 days; the latest day or two is preliminary and shows as a dotted line. An Hourly view exists in the 24-hour view only.

When the Discover report appears, and why it can be missing

Google’s Discover performance report help page says the report shows only after a property reaches a minimum number of Discover impressions. Google doesn’t publish the number, so treat any figure quoted elsewhere as a guess. If Discover isn’t listed under Performance, check these causes in order:

1. Not enough Discover impressions yet. The usual reason for small or new sites. There is nothing to fix.
2. You’re in the wrong property. Data shows only in the property that contains the canonical URL. If your canonicals use https://www. and you opened a URL-prefix property for the non-www version, you’ll see nothing. A Domain property avoids this; see Domain vs URL-prefix property.
3. Your articles aren’t indexed. Discover only shows indexed content. Check recent articles with URL Inspection, or ask your assistant to run inspect_url on them.
4. A policy problem. Open Security & Manual actions → Manual actions. Discover manual actions appear there, and repeated or egregious violations can cost a site its Discover eligibility.

Once the report exists, low-impression pages and countries get no row of their own but still count in the totals, so the table may not add up to the chart. On small sites, low days can be left off the chart too: a missing day isn’t necessarily a zero day.

Discover vs Search: the differences that matter

Discover sits next to the Search results report but works differently in almost every way that affects how you read the numbers.

The difference that trips people up most: the Search results report never includes Discover. Google’s Search Analytics API reference says the web search type excludes Discover and Google News results, so a Discover collapse never shows in your Search totals. If analytics shows fewer Google visits while Search is flat, look at Discover first. For how Search defines its own metrics, see Search Console metrics explained.

Discover and Google News data in the Search Console API

The Search Analytics API has served Discover and Google News data since October 2021 through the type parameter (formerly searchType, which still works). Google’s announcement said plainly that “queries and positions are not supported by the Google Discover report”, and that requesting an unsupported dimension returns an error. Google’s performance data documentation adds that position is not recorded for Discover or Google News.

Don’t confuse news with googleNews: the first is a tab inside Google Search, the second is the separate Google News product, whose report does have a Device tab.

Dimensions that work with Discover

Use page (the canonical URL), date (Pacific Time days), country (three-letter codes such as usa) and searchAppearance, whose Discover values are AMP_TOP_STORIES (“AMP articles”), NEWS_SHOWCASE, AMP_STORY (Web Stories) and VIDEO. Leave out query, as a dimension and as a filter, and ignore any position value a tool shows you. Google doesn’t document device for Discover and the report has no Devices tab, so don’t rely on it. Page filters work normally, including includingRegex for splitting a site into sections.

Which GSC MCP Server tools read Discover

Only query_search_analytics: set searchType to discover (or googleNews), use up to four dimensions and 10 filters, and page through results 1,000 rows at a time with startRow. Our shortcut tools, such as get_search_overview and compare_search_periods, read web search (and Bing) only. To compare two Discover periods, ask for one query_search_analytics call per period, both inside one multicall. Our Google Search Console API guide covers row limits, freshness and quotas.

How to get on Google Discover

According to Google’s Discover documentation (Google’s posts call it the “Get on Discover” page), content is eligible automatically if Google has indexed it and it meets Discover’s content policies. No special tags, structured data or submission are needed, and eligibility doesn’t guarantee appearing.

Google’s recommendations

• No clickbait or sensationalism. No misleading or exaggerated titles, snippets or images, no withholding what the piece is about, and no appeals to morbid curiosity, titillation or outrage.
• Titles that capture the essence of the content, describing the article rather than teasing it.
• Content that’s timely, tells a story well or offers unique insights.
• Large, high-quality images: at least 1,200 pixels wide, more than 300,000 pixels in total, ideally 16:9, with large previews enabled by max-image-preview:large or AMP.
• A preferred image in schema.org markup (such as primaryImageOfPage) or og:image: relevant and representative, not your logo and not text-heavy. Google added this preferred image guidance to its Discover documentation in March 2026.
• A good page experience overall.

Expertise is judged topic by topic

Discover uses many of the same signals and systems as Search to decide what is helpful, people-first content, so Google points Discover publishers to its guidance on helpful, reliable, people-first content, the page that explains E-E-A-T (experience, expertise, authoritativeness and trustworthiness). The February 2026 Discover core update went further: Google wants more in-depth, original and timely content from sites with expertise in a given area, judged topic by topic. A local news site with a real gardening section can count as expert on gardening; a film review site with one gardening article probably won’t.

Content policies and transparency

Discover’s own content policies sit on top of Google Search’s spam policies and cover categories such as dangerous, deceptive, hateful, sexually explicit and violent content, manipulated media and medical content. Previews mustn’t promise what the page doesn’t deliver, sponsored content must be disclosed, and news sources should show dates, bylines, author and publisher details and contact information. Discover also filters out things like job applications, petitions and forms.

Discover changes in 2025 and 2026

Discover has changed several times since September 2025. These are the changes we could confirm on Google’s own channels.

Two things remain unconfirmed. Google said the Discover core update would expand to all countries and languages in the following months; we found no announcement that this is complete (checked on 30 October 2026). And Discover on desktop appeared on google.com in some countries in 2025, but Search Engine Roundtable reported in September 2026 that it had not been seen for months, and Google hasn’t commented. Treat Discover as a mobile surface for now.

Why Discover traffic is so volatile

Google’s documentation tells site owners to treat Discover traffic as supplemental to keyword-driven Search traffic, and gives three reasons it moves:

• Interests change. When people lose interest in a topic, partly reflected in their searches, feeds show it less.
• Content types change. Google keeps adjusting which kinds of content the feed shows.
• Search updates. Core updates can move Discover, and since February 2026 it has had its own. Google adds that some changes are unrelated to your content’s quality or publishing frequency, and sometimes there is nothing to do.

The mechanics make it spikier still. Search traffic has a floor: people keep searching for how to descale a kettle whether or not you published anything this week. Discover has no such floor. An article gets its chance while it’s new and relevant to people’s interests, so your Discover total is largely the sum of recent bursts. If a handful of pages carry most of it, one quiet week can look like a collapse. Check how concentrated yours is:

That second answer is your typical Discover lifespan. If pages earn nearly everything within days, weekly swings mostly reflect what you published, not what Google did.

Google Discover traffic drop or Search drop?

First find out which surface fell. A Search drop is about rankings, indexing and results pages, which our organic traffic drop diagnosis covers step by step. A Discover drop is about content, eligibility and feeds. Work through these checks in order:

1. Split Google clicks by surface over the same period and see which line moved, and from which day.
2. Rule out data problems on Google’s data anomalies page. Each Discover logging error in the timeline above lowered one or two days without any real traffic loss.
3. Check for updates. The Search Status Dashboard lists core updates starting 11 December 2025, 27 March 2026 and 21 May 2026, plus the 5 February Discover update. 21 May was also a Discover logging error day, so judge that update on the days after.
4. Find out whether it’s a few pages or everything. If last month’s three biggest hits aged out and nothing replaced them, that is a content pipeline gap, not a penalty.
5. Check eligibility: Manual actions, inspect_url on recent articles for index status and canonical, and whether your template still outputs max-image-preview:large and a large og:image.

A slow slide on evergreen pages in Search is a different problem again: see the content decay audit.

A weekly Discover analysis workflow with AI

Discover moves weekly, so review it weekly. Use Search Console’s own weeks (Sunday to Saturday, Pacific Time) and complete weeks only: run on 30 October 2026, that means 18–24 October against 11–17 October. Once your assistant is connected (see how to connect Search Console to Claude, ChatGPT and Cursor), the routine has five parts:

1. Totals by surface: daily Discover and web clicks for both weeks, so you know whether this week’s story is Discover, Search or both.
2. Pages, week against week: new winners, pages that faded, and pages whose CTR sits well below their section’s median.
3. Sections and topics: group pages by the first URL folder, or by a regex per section such as /recipes/. Topic-level results matter now that Google judges expertise topic by topic. The Search Console regex cheat sheet has tested page patterns.
4. The misses: run inspect_url on two or three articles you expected to do well that got no Discover impressions. An unindexed page or an unexpected canonical is a technical problem, not an editorial one.
5. The note: three lines for editors: what earned Discover, what to publish more of, and anything that looks broken.

That prompt costs four tool calls, because each call inside a multicall counts, and the inspect_url checks bring a week to about seven. Weekly, that is roughly the Free plan’s 30 calls a month for one site; Pro raises it to 10,000 calls across five accounts (see pricing). Agencies can reuse the same pulls in an AI client reporting workflow.

Google Discover optimization: from data to decisions

The weekly numbers only matter if they change what you publish.

Back the topics that already earn Discover

If two sections earn most of your Discover clicks, that is where Google’s systems see your expertise. Deepening them is a better bet than chasing a viral hit in new topics, and fits the February 2026 update’s focus on topic-level expertise.

Fix previews before rewriting articles

In Discover the preview is the title and image, so plenty of impressions with a weak CTR usually means a preview problem. Compare CTR within a section and format (the Discover appearance grouping separates videos and Web Stories from articles), then check whether the headline describes the piece and the image is large and relevant. Don’t answer a low CTR with a more sensational headline: the 2026 update was designed to show less of that.

Audit image tags in your templates

In Claude Code, Cursor or VS Code, ask the assistant to read your article template and confirm it outputs max-image-preview:large, an og:image at least 1,200 pixels wide, and a representative image in the structured data. Google says large images are more likely to generate Discover visits, so a template without them holds back every article built on it. Our guide to Claude Code for SEO covers the setup.

Keep useful older articles current, and don’t chase spikes

Google says older content can appear when it’s relevant to someone’s interests, so an accurate, updated explainer has a chance when its topic returns to the news. Update the substance, not just the date. And judge patterns over several weeks: one article with 50,000 Discover clicks tells you little on its own.

What Discover data can’t tell you

• Why someone saw your card. No queries, positions or interest data; infer it from the topics that perform.
• Devices. No Devices tab, and no documented device split for Discover.
• What happened after the click. Reading time, subscriptions and revenue need analytics, and Google Analytics has no Discover channel. See Search Console vs Google Analytics 4 for lining the two up.
• AI features in Discover. Those impressions have a separate, impressions-only generative AI report that our tools can’t read; see Search Console’s generative AI report.
• Manual actions. These appear only in the Search Console interface, which our read-only server can’t see, so check them yourself after any sharp fall.
• Anything older than 16 months. Export monthly, or set up Search Console’s bulk data export to BigQuery, which keeps data from the day you start it.

Within those limits, the Discover report is the only first-party record of how your content does in Google’s feed: pull it weekly, read it by section, and judge changes over several weeks.

Frequently asked questions