Marketing Intelligence Weekly: Search Checks, Ad Delivery and the Numbers Behind Reach
Date: 3rd October 2026
TL;DR
Google's September spam update is still rolling out, while its revised guidance puts explicit emphasis on manually checking AI-generated content. Neither announcement proves that a particular site's traffic loss was caused by AI copy.
Google has opened AI Mode's information-monitoring feature more widely. It is a new way for people to follow changing subjects, not yet a demonstrated source of brand referrals.
Performance Max account shows why teams should inspect targets and traffic mix after Google's August bidding change, rather than assume a platform-wide decline from a single result.
Shopify changed how it counts sessions and filters identified bots. A conversion-rate movement around the change may reflect the denominator, not a change in orders.
Search, Social & Platform Updates
Google's spam update calls for diagnosis, not a quick rewrite.
The September 2026 spam update began on 24 September and applies globally, across languages. Google said the rollout could take up to two weeks. One operator of a Madrid events calendar reported losing roughly 95% of Google Search visits overnight soon afterwards, despite finding no obvious robots, canonical or sitemap problem. That is a serious account to investigate, not proof that the update caused the fall or that Google has singled out event calendars.
For sites watching the same rollout, the first useful distinction is whether a loss is concentrated in particular queries and pages or appears across the whole site. Check Search Console alongside independent analytics, annotate the rollout date, and look for indexing or technical changes before rewriting pages at scale. A traffic chart alone cannot identify the cause.
AI Mode can now keep watch on a subject for the searcher.
Google said it was expanding AI Mode's information-monitoring capability beyond its paid tiers. A user can describe what to watch and receive updates as information changes across the web. A marketer in this week's discussion proposed asking it to follow a topic's Reddit conversations. That is an interesting discovery behaviour, but it does not establish that brands can buy placement, control the alerts or expect traffic from them.
Why it matters
Search is changing in two different ways: rankings can move during a broad spam rollout, and people can ask an AI interface to keep searching over time. The response to both is better observation, not a hurried “AI SEO” tactic.
What you can do
Mark 24 September on Search reports. Compare affected and unaffected pages and queries, then check crawl, indexing and recent site changes before assigning a cause.
Try AI Mode monitoring for a topic your audience follows. Record what sources it surfaces and whether your content is useful in that context; do not count an alert as a referral.
Paid Media & Acquisition
Performance Max deserves a target-and-mix audit.
Google's 17 August bidding change affects target-based campaigns marked “Limited by budget”, including Performance Max. It is designed to bring delivery closer to the target the advertiser sets. One ecommerce operator said they raised target ROAS from about 450% to 700% ahead of the change. Orders then fell about 28% against their June–August baseline. Their eight strongest product pages fell from 22.9% to 16.6% of Google CPC sessions, while generic search-term spend rose from 23% to 30% in the periods they compared.
That is unusually useful account detail, but it cannot isolate one cause. The target itself changed, the comparison spans different periods, and the operator had only five days of data after lowering it again. The lesson is to inspect the mix that produced the result: products, queries, channels, spend and qualified orders. Do not rebuild a campaign because another account's graph looks familiar.
The product feed is part of acquisition, too.
Google confirms that product data helps it understand and show listings, and that custom labels can organise Shopping campaigns. The post offers a practical audit list, not measured proof that any one feed edit will lift return on ad spend.
Why it matters
An automated campaign can meet its configured target while sending money towards a different set of products or searches. Better bids will not compensate for a feed that describes the wrong item or hides the margin of the one it promotes.
What you can do
For budget-limited campaigns, record target changes alongside product, query and channel mix. Compare orders and contribution margin, not ROAS alone.
Audit a small group of high-value products for accurate titles, identifiers and margin labels. Change one feed issue at a time and keep a clean comparison period.
Content, Brand & Communications
Google has made the human review step explicit for AI content.
Its updated Search guidance says publishers should manually check generated material for accuracy and trustworthiness before publication. That extends beyond the article body to titles, descriptions, structured data and image alt text. This is not an announcement of a blanket ban on AI writing or a new penalty for using it. It is a reminder that a polished-looking page can still misstate facts in the parts a reader or search result sees first.
Without the site's history and Search data, nobody can know whether AI use caused the decline. A better editorial response is to identify pages that are inaccurate, repetitive or unhelpful and improve or remove them for readers, rather than treating a model-generated label as the diagnosis.
Why it matters
Quality control and audience interest are separate jobs. Fact-checking keeps a claim from being wrong; a specific, relevant premise gives someone a reason to care.
What you can do
Add a pre-publication check for claims and metadata in AI-assisted pieces, assigning a human owner to every factual assertion.
Test different premises on the same audience before investing in more production. Compare early retention and meaningful replies, not views in isolation.
Measurement & MarTech
Shopify's session number changed, not necessarily your customer count.
Between 21 and 23 September, Shopify changed session measurement and began excluding identified bot sessions from session-based reports by default. It also counts some sessions without a page view and no longer splits continuing activity at midnight UTC. Depending on the store, sessions may rise or fall; conversion rate can move because its denominator changed. Shopify says total orders, sales and customer counts are not affected by this measurement update.
A highly upvoted Meta Ads post treated the change as proof that Meta knowingly charged advertisers for bot clicks. Shopify's documentation does not establish that accusation. Shopify sessions and Meta-reported ad clicks are different measures, and the update changes how one of them is counted. It gives teams a reason to reconcile dashboards, not a verdict on another platform's intent.
More views don’t always mean more customers.
A campaign can generate thousands or even millions of views and still have little impact on actual sales. On the other hand, more relevant and engaged audience may drive stronger conversions.That’s why influencer performance shouldn’t be judged by reach alone. When evaluating a campaign, look at the bigger picture: how many people took action, how many became customers, what the campaign cost, and what revenue it generated. Views and impressions are useful indicators of attention, but they don’t necessarily show business impact.
Why it matters
This week's most expensive mistake would be mistaking a changed metric for a changed business. Reach, sessions, clicks and sales answer different questions.
What you can do
Put the Shopify rollout date in ecommerce reports and set a new baseline for session-based rates. Check the human/bot filter and compare orders and revenue separately.
Give each creator a trackable offer or link, record the cost and measure attributable sales over an agreed window. Treat views as context, not the final result.
Comments