Currently Available: Need a skilled Software Developer for your next project?
Categories
SEO, GEO and AEO

Querying Search Console Image and Video Data in BigQuery

Search Console’s BigQuery export lets website owners use SQL to analyze performance data from Google Search. For image and video reports, choose which Search Console search type to query: image includes traffic from the Image tab, and video includes traffic from the Video tab. These values identify the tab where a search happened. They do not say whether a Web result contains an image or video.

Filter for the search type you want to report, then choose a table that fits your question. Group rows at the level you need, and sum clicks, impressions, and position values before calculating click-through rate or average position. The examples below show how to keep image or video results separate from ordinary Web reporting.

Filter by search type

Use the search_type field to separate Image-tab or Video-tab traffic from Web search. For a report that compares both, include search_type in the output and group by it. For an Image-only report, filter to image; for a Video-only report, filter to video. Google identifies these as the Image and Video search types in its BigQuery table reference.

A search appearance describes a feature shown in a result, such as a video feature in Web results. It answers a different question from search type, which identifies the Search Console tab or area where the search happened. Filtering for a video appearance does not produce the same report as filtering for the Video search type.

The Search Console Performance report added a Web multimodal category for searches involving experiences such as Lens and Circle to Search. Google describes this category separately from the Image tab. The reviewed BigQuery schema, which describes the exported data fields, does not explain how the category appears in the export. Treat it as a separate category in your analysis, and check how BigQuery represents it rather than assuming an Image filter includes it. Google’s multimodal reporting announcement describes the new category.

Choose a table for the question

Use the searchdata_site_impression table for reports about a Search Console property, such as daily Image and Video trends or query breakdowns. Use the searchdata_url_impression table for reports about individual URLs, such as which pages received traffic. Both tables can contain repeated dimension keys, meaning the same dimension values appear in multiple rows. Sum the metric fields when you group those rows.

For example, this query produces a daily trend with separate rows for Image and Video search:

SELECT
  data_date,
  LOWER(search_type) AS search_type,
  SUM(impressions) AS impressions,
  SUM(clicks) AS clicks,
  SAFE_DIVIDE(SUM(clicks), SUM(impressions)) AS ctr,
  SAFE_DIVIDE(SUM(sum_top_position), SUM(impressions)) + 1 AS avg_position
FROM <code>project.dataset.searchdata_site_impression
WHERE data_date BETWEEN DATE '2026-10-01' AND DATE '2026-10-31'
  AND LOWER(search_type) IN ('image', 'video')
GROUP BY data_date, search_type
ORDER BY data_date, search_type;

Replace the project and dataset names with the names in your BigQuery export. The date filter also limits how much date-partitioned data BigQuery scans. Google’s query guidance recommends restricting the date range for this reason.

Use the URL table to find which landing pages receive Image-tab traffic. In Search Console’s image reporting, the URL identifies the page Google associates with the image result, called the canonical host page. It does not identify the image file shown, and the export does not distinguish between multiple images on the same page. This query groups Image-tab activity by the URL in the report:

SELECT
  url,
  SUM(impressions) AS impressions,
  SUM(clicks) AS clicks,
  SAFE_DIVIDE(SUM(clicks), SUM(impressions)) AS ctr,
  SAFE_DIVIDE(SUM(sum_position), SUM(impressions)) + 1 AS avg_position
FROM <code>project.dataset.searchdata_url_impression
WHERE data_date BETWEEN DATE '2026-10-01' AND DATE '2026-10-31'
  AND LOWER(search_type) = 'image'
GROUP BY url
ORDER BY clicks DESC;

The property and URL tables use different fields for position. Use sum_top_position in the property table and sum_position in the URL table. Check the field names in your exported tables before running a query because the documented schema uses different names in some places.

Calculate metrics after grouping

The fields in GROUP BY determine what each output row represents. Grouping by date gives a daily trend. Adding url gives a page breakdown, and adding query gives a query breakdown. Each added field makes the output more detailed and changes what its totals mean.

Before calculating other metrics, sum clicks and impressions, along with the relevant position field. Calculate click-through rate (CTR) by dividing total clicks by total impressions. Calculate average position by dividing the position sum by total impressions, then adding one. Search Console counts positions from zero, so the first position has a value of zero. Adding one puts the result on the average-position scale shown in Search Console. Google explains these metrics in its clicks, impressions, and position guidance.

Do not average row-level CTRs or average positions to produce a larger-group result. A row with few impressions would otherwise count as much as a row with many impressions. The query examples avoid that error by summing the underlying metrics first and deriving the ratios from those sums.

Keep query reports separate from total reports

A query breakdown lists search queries by name. A total-performance report counts metric rows even when query text is unavailable. Search Console marks anonymized queries with is_anonymized_query, and their query text is blank. Exclude those rows when ranking named queries. But do not use that filter in a total-clicks or total-impressions query unless you intend to leave out rows with anonymized query text.

This example reports named Image-tab queries:

SELECT
  query,
  SUM(impressions) AS impressions,
  SUM(clicks) AS clicks
FROM <code>project.dataset.searchdata_site_impression
WHERE data_date BETWEEN DATE '2026-10-01' AND DATE '2026-10-31'
  AND LOWER(search_type) = 'image'
  AND is_anonymized_query = FALSE
  AND query != ''
GROUP BY query
ORDER BY clicks DESC;

The blank-query condition is appropriate here because every output row needs a query label. It is not appropriate for an all-activity total. Google’s bulk export overview and schema guidance describe anonymized-query data in different ways, so avoid assuming that a blank-query filter preserves every metric total.

Avoid misleading comparisons

If you leave out search_type, the report combines categories and answers a different question. For example, a total grouped only by date combines Image, Video, and other search types in the selected table. Filter on search_type to report one category, or group by search_type to compare categories.

Label totals for each search type separately. Google records Image and Web traffic separately, and the same image can appear in both. Adding the totals gives a sum across Search Console search types, not a count of unique people, searches, or visits. Image-tab metrics do not count every image interaction as a click. Google’s metric definitions explain how clicks and impressions are counted for image results. Do not apply those image-specific details to video reporting without evidence that the same rules apply.

Do not join the site and URL tables before you aggregate them. The tables contain different levels of detail, so a join can multiply matching rows and inflate metrics. First aggregate each table to a compatible level. Then join them only if your reporting question requires it. When comparing BigQuery totals with Search Console, match the property and date range. Use the same search type and grouping in both reports. A query report for a whole property and a total rolled up from individual URLs do not necessarily match.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com

Email updates

Usually a new article and a few links I found interesting.

No spam. Unsubscribe with one click.

Leave a Reply

Your email address will not be published. Required fields are marked *