Search digestSeptember 2026

September's spam rollout, AI and visual search reports, crawler controls and Yandex review rewards — with evidence and practical site checks.

Issue 1010 min read18 primary sources

September ended with a fresh spam rollout and several useful ways to inspect how sites appear in search. The work splits into three decisions: investigate a ranking loss, measure exposure that used to be hidden, and check eligibility for a specific search feature. Changes to a dashboard or a guide should not be treated as evidence that every site's rankings changed.

September 8Regional search-feature guide

Eligibility and regional requirements become easier to inspect.

September 15Cloudflare crawler controls

Disallowing training and blocking access become distinct choices.

September 22 / 29Microsoft Clarity reporting controls

Page classifications and custom brand terms improve report definitions.

September 24Spam rollout and multimodal reporting

A ranking rollout and a new measurement filter require separate analysis.

Product change

A new spam rollout began before the month closed

Google released the September 2026 spam update on 24 September at 09:15 Pacific time. The incident record says it applies globally and to all languages, with a rollout that may take up to two weeks. It contains no completion notice as checked on 7 October. September therefore closes with an announced rollout, not a measured final verdict.

The ranking history records no separate September core update. Keep that distinction in reporting: this is a spam update, and a movement in the same window does not establish why an individual page moved.

Google's spam-update guidance directs affected owners to review its spam policies. It also says improvement can take months as automated systems learn that a site complies. That is a reason to inspect the affected page groups and actual policy issues before promising a quick recovery.

What to do: mark 24 September in your reporting, retain a pre-rollout baseline, and compare the same pages, queries and countries. Separate observed losses from a hypothesis about their cause. Do not reuse August's impact percentages as September measurements.

Product change

Cloudflare separates refusing training from blocking search

Cloudflare's 15 September crawler-control release adds Disallow AI Training. It publishes preferences through Bot Preference Sync while allowing qualifying mixed-use crawlers to continue serving search. Crucially, Training → Block and Block on pages with ads now also apply to mixed-use crawlers such as Googlebot, Bingbot and Applebot. The stronger setting can therefore affect search discoverability.

Existing preferences migrate according to Cloudflare's published rules; a legacy training block should not be assumed to have the new meaning. Inspect the effective Search, Training and Agent settings, the served robots.txt and actual allowed crawler requests after migration. A firewall decision and a robots.txt preference are different layers.

Before choosing a setting: decide whether the goal is to refuse training or stop access altogether. Cloudflare says Bing's robots.txt training opt-out is targeted for early 2027, so Disallow AI Training does not yet automatically convey that preference to Bing. Do not report it as a complete cross-provider opt-out.
Cloudflare's September controls distinguish purpose and access. Operator support still differs; a training preference is not the same as blocking the request.

Product change

Search Console makes visual search easier to isolate

On 24 September, Google introduced web multimodal search reporting in both Search results performance and Generative AI performance reports. The new search-type filter covers Lens, Circle to Search on Android, images uploaded to Google Search, and Chrome's right-click image search. The global rollout starts on that date; metrics appear when a site receives traffic from those queries.

Use the filter and export the result to establish a separate baseline for pages discovered through visual input. For a retailer, image-led catalogue or tutorial publisher, compare the pages receiving that exposure with their actual visits and business outcomes. The announcement describes UI filtering and exports; it does not announce a new Search Analytics API parameter.

Reading the report: an empty view is not proof that your images are unindexable. Check rollout availability and whether the property receives qualifying traffic before treating an absence of metrics as a technical fault.

Product change

Retailers gain AI visibility insights in five markets

Google's 16 September commerce announcement makes AI performance insights in Merchant Center generally available in Australia, Canada, India, New Zealand and the United States. It describes brand and product share-of-voice comparisons across surfaces including AI Mode and AI Overviews.

The same announcement adds cart transfer to a merchant site and enhanced checkout-flow testing to the Universal Commerce Protocol integration hub. Those capabilities are rolling out gradually in the U.S.; Australia and Canada are planned for early 2027. Treat that timetable separately from the five-country availability of AI insights.

For an eligible merchant, check whether the reports expose useful differences between products and brands, then compare them with sales and referral data. If the business uses UCP checkout, test a real cart handoff and the resulting order flow. A visibility metric cannot substitute for successful checkout or revenue attribution.

Data

French CTR research gives a reason to inspect clicks

An Ahrefs study published on 16 September analysed 963 domains using aggregated Search Console data: 28 days before France's 22 July AI Overviews launch and nine days after. Its high-exposure group had AI Overviews on more than 20% of French queries; the control group had less than 2% exposure.

Group / measureReported change
High-exposure group: raw median CTR−23.1% relative
Low-exposure control group: CTR+2.3% relative
Ahrefs' control-adjusted estimate−24.8% relative

These are relative changes, not percentage-point losses. The study covers a selected French domain panel and a short July window; it is September research, not a September rollout. Its observational comparison is not a universal forecast or proof of the cause of every site's decline.

Use it to choose a check: compare clicks, CTR and stable query groups in each important market. Rising impressions can coexist with weaker click-through. Keep actual revenue alongside those search metrics before changing content investment.
Ahrefs' raw group changes, with a zero reference. The control-adjusted estimate is a separate calculation; this short observational panel is not a forecast for every site.

Product change

Clarity adds two useful checks for AI measurement

On 22 September, Page Classification arrived in Clarity's Bot Activity dashboard. It combines content-based classification and URL rules to group pages by type. Owners can create categories or add paths to existing ones; custom rules take precedence over automatic classification.

For a site with mixed or irregular URLs, inspect several examples in each category before comparing product, editorial and contact pages. A useful page-group view starts with correct membership. Bot requests show activity on those pages; by themselves they do not establish a citation, a human visit or a conversion.

On 29 September, custom Brand Terms added another measurement control. In Settings → AI Visibility, teams can edit the pre-populated terms, group variants under a Brand Name and filter by either the name or an individual term. Abbreviations, localised names and common misspellings can make branded/non-branded classification more accurate.

Record classification changes: save the date and terms or page rules you edit. If the branded share or a page group's activity changes afterwards, check whether the definition changed before claiming new demand. Avoid adding ambiguous generic words to a brand group merely to improve its apparent visibility.
Editorial diagram: each report answers a different question. A crawl request does not demonstrate a citation or a successful sale.

Product change

Local-business search opportunities depend on the market

Google added a regional search-feature guide on 8 September. On 18 September, its documentation changelog recorded local-business query support for aggregator and supplier units. This is an eligibility and implementation story for relevant businesses, not a worldwide ranking-system update.

In the EEA, the aggregator unit is for approved Vertical Search Services such as directories and comparison providers. Participation involves relevant content, data and Google's quality requirements; an interest form is a starting point, not automatic inclusion.

The separate supplier unit serves direct providers and appears only alongside an aggregator unit. Google says no additional data beyond what crawling can access is required, though feeds can enhance results. First identify the business's role, market and query eligibility. A directory and a local business should not receive the same integration plan.

Scope the work: check the intended country, query category, provider role and participation requirements before assigning markup or feed development. The regional guide also covers other features; availability of one does not establish eligibility for another.

Product change

Search profiles become more practical for U.S. publishers

Google's 16 September Search profile update adds management of multiple brands from one login, revised article thumbnails and longer headlines. It also lowers the eligibility threshold for more U.S. publishers and creators to 10,000 followers across YouTube, Instagram, X or TikTok. The announced expansion is U.S.-specific.

The accompanying badge guide explains how to link a website to a claimed Search profile. Google says following that profile makes its linked content more likely to appear to that audience in Discover. A text link is also supported; the badge is not the same mechanism as Preferred Sources.

For an eligible publisher, verify the claimed profile, linked accounts and article presentation before placing an invitation on the site. Track whether readers use it. Do not promise a general ranking uplift or assume that the U.S. access threshold applies worldwide.

Product change

Two documentation changes deserve narrow implementation checks

The 24 September VideoObject documentation update documents support for creator, including author as an alternative, and clarifies interaction types. The video guide accepts a Person or Organization, with name or alternateName, and lists WatchAction, LikeAction, CommentAction and ShareAction for interaction statistics.

If your video template supplies these fields, validate a representative page and ensure each count describes its stated action. Do not put the same total under several interaction types. This is documentation of supported properties, not evidence that every video page needs a new markup field or that eligibility guarantees a rich result.

On 17 September, Google's crawling changelog clarified that rules addressed to Mediapartners-Google affect several advertising products, including AdSense and Ad Manager. Review narrowly scoped robots.txt rules if the site uses those products. The change clarifies scope; it does not announce a new crawler or a Google Search ranking change.

Product change

Yandex changes review rewards for its regional market

On 7 September, Yandex updated Points for Reviews. Participating ecommerce and service sites can choose rewards of 40, 100, 150 or 200 Plus points, replacing the single 40-point option. The setting lives in Webmaster's Reviews and Rating in Search section.

Yandex says a new reward takes effect within two days; reviews already under moderation retain the previous reward. For a business using that ecosystem, this is a change to customer-feedback collection and its budget. It is region-specific, and the announcement does not establish a ranking uplift or a new search algorithm.

Owner action: check participation requirements, the current reward and the cost of collecting useful feedback before changing the setting. Review volume, the quality of feedback and search performance should be measured separately.

Action

What to check on your own site

  • A late-September visibility loss: mark the spam rollout, compare consistent page/query/country groups and inspect concrete spam-policy issues before changing the whole site.
  • A Cloudflare-hosted site: inspect the migrated Search, Training and Agent controls, served robots.txt and permitted crawler requests before changing a training preference.
  • A site with important visual content: export the new multimodal view and save its first useful baseline, keeping exposure separate from referrals and conversions.
  • An eligible retailer: inspect Merchant Center AI insights and test any UCP cart transfer end to end.
  • An AI measurement workflow: verify Clarity's page classifications and brand variants; record edits so reporting comparisons retain their meaning.
  • A local provider, directory or publisher: verify region-specific eligibility before commissioning a feature integration or adding a Search profile invitation.
  • A video or advertising template: validate the relevant fields and crawler rules on representative pages; avoid a broad rewrite for a documentation-only clarification.
  • A business using Yandex reviews: check the new reward choices and their budget; do not equate paying for feedback with a guaranteed ranking gain.

Sources