Gemini 3.8 AI Mode and Search Console's AI Reports: What E-commerce Stores Can See and Do Now

Gemini 3.8 AI Mode is live for paying subscribers, Search Console now reports AI impressions for every site, and an AI contribution pilot is in testing. What each means for e-commerce visibility.

Gemini 3.8 AI Mode is the newest combination in Google Search: since 2 September 2026, Google's Gemini 3.8 Flash model has been available inside AI Mode for Google AI Pro and Ultra subscribers. Alongside it, Search Console has added a generative AI performance report that, since 31 August, every website can use to see how often its links appear in AI Overviews and AI Mode, and Google is piloting something called AI contribution with a small group of publishers. For an online store, these three developments answer different questions: what is generating the AI answers, how often your pages appear in them, and whether Google might one day pay for content that feeds them. Only the first two are available to everyone today.

We build software and storefront technology for Shopify merchants, and the question we are asked most about AI search is the practical one: can I see whether it is sending me anything, and what should I do about it? This post sets out what Google has actually announced, with sources, what is only a test, and what an e-commerce team can measure and change this month.

What Gemini 3.8 Flash is, according to Google

Google introduced Gemini 3.8 Flash on its Keyword blog on 2 September 2026, in a post by Tulsee Doshi, Senior Director of Product Management, and Raluca Ada Popa of Google DeepMind. Google described the model as delivering significant improvements over Gemini 3.7 Flash across software engineering, agentic tasks and multi-step reasoning in specialised domains. The same post announced a cybersecurity variant, 3.8 Flash Cyber.

On availability, Google's post says 3.8 Flash is available to Google AI Pro and Ultra subscribers across the Gemini app, AI Mode in Google Search and Gemini in Google Sheets, and to developers through Google AI Studio, Android Studio and the Gemini API. Robby Stein, VP of Product at Google Search, announced the Search side of the launch, and Nick Fox, who leads Google Search, noted that it arrived less than three weeks after Gemini 3.7 Flash. Search Engine Roundtable reported that subscribers select it from the model menu in AI Mode.

Two points matter for a store owner. First, this is a model available to paying subscribers who choose it. Google has not announced that Gemini 3.8 Flash is the default model for everyone using AI Mode for free, and we have not seen anything from Google saying so. Second, model upgrades in AI Mode are now arriving weeks apart. Any conclusion drawn from how one model answers shopping questions has a short shelf life.

The citation bug, and why it matters for Gemini 3.8 AI Mode

Within a day of the launch, Search Engine Roundtable reported that AI Mode answers generated with Gemini 3.8 Flash were showing far fewer links and citations to sources than before, with examples posted by SEO consultant Gagan Ghotra. On 4 September, Robby Stein acknowledged it on X, saying it was not working as intended and that a fix would roll out soon. By the next morning, SEO consultant Glenn Gabe reported that 3.8 Flash was providing links again.

The episode was short, but it shows something worth remembering. Whether an AI answer sends a visitor to your site depends not only on your content but on how a particular model, on a particular day, decides to cite sources. Google treated missing citations as a bug, which is a useful signal of intent. It also means a sudden dip in AI-driven visits is not always about your site.

Search Console's generative AI performance report

The more practical development for most stores is in Search Console. Google announced generative AI performance reports on the Search Central blog in June 2026, and Search Console Help says that as of 31 August 2026 the insights had rolled out to all websites worldwide. Here is what the Search report shows, according to Google's Help page.

• Impressions, defined as how many times links to your site were shown to a user in a generative AI feature on Google Search. If several links to your site appear in the same AI feature, they count as one impression.
• AI Overviews and AI Mode together. The report combines impressions from both features rather than separating them.
• Breakdowns by page (the canonical URL), country, date and device.
• A search type filter that separates text-based web search from multimodal search, such as Lens, Circle to Search and searches with uploaded images.
• A table limited to 1,000 rows, with the newest data marked as preliminary.
• No data from Search Labs experiments, and pages need enough impressions to appear.

What it does not show is clicks. Search Engine Journal's report on the worldwide rollout noted that click data is not available in these reports. Google's separate documentation on AI features says clicks and impressions from AI features are included in the main Performance report under the Web search type, so the clicks are in Search Console, but not separated out as AI clicks.

Search Engine Journal also reported a property-level control in Search Console that lets a site opt out of AI Overviews, AI Mode and Discover's generative AI features, and quoted Google as saying the setting is not used as a ranking signal elsewhere in Search and has no effect on AI training. We would treat that as a decision for the business, not for an SEO tool, and for most stores we would not recommend opting out without first measuring what AI features are doing for you.

The AI contribution pilot: a test, not a feature

The third development is the one most likely to be misreported, so we will be precise. In April 2026, Search Engine Roundtable noticed references to an 'AI contribution' report in Search Console's support documentation, with no screenshots or public details. Google's John Mueller gave only a general comment at the time.

In mid-September, Search Engine Roundtable and 9to5Google, citing reporting by Digiday, described what it turned out to be: an AI contribution pilot in which Google pays some publishers when their content contributes significantly to AI-generated responses in Gemini, AI Overviews and AI Mode. Participants see an AI earnings widget in Search Console with a monthly earnings figure and history. Digiday reported that at least dozens of publishers had been approached, mostly small and mid-sized. Google described it as an early-stage learning pilot to test how best to reward high-quality content, on top of the traffic and tools it already provides.

For an online store, the honest summary is: this is a limited pilot aimed at publishers, Google has not published eligibility rules or how payments are calculated, and there is nothing a merchant can apply for or optimise towards today. We have seen claims about how many publishers are in it and what share of revenue it pays; we could not verify them at a primary source, so we have left them out. If the pilot ever widens to commerce content, it will be a significant change. Until then, it is something to watch, not something to plan around.

What Google says about appearing in AI features

Google's documentation on AI features and your website, last updated in December 2025, is short and consistent with everything above. Its main points for a site owner are these.

• There are no special optimisations or additional technical requirements to appear in AI Overviews or AI Mode. A page must be indexed and eligible to be shown in Google Search with a snippet.
• The usual fundamentals apply: crawlable pages, sensible internal linking, good page experience and content that matches what people are looking for.
• AI Mode and AI Overviews can use a technique Google calls query fan-out, issuing several related searches to build a response, which can surface a wider set of links than a traditional results page.
• Existing controls, such as nosnippet, data-nosnippet, max-snippet and noindex, govern how content can appear.
• Traffic from AI features is reported within Search Console's Performance report under the Web search type.

That is less exciting than much of what is written about AI search, but it is useful. It means the work that gets a store into AI answers is mostly the work that gets it into ordinary search results: pages Google can crawl, index and quote, with content that is specific and accurate.

What this means for e-commerce visibility

Put together, the three developments suggest a few practical conclusions for stores.

• You can now measure AI visibility, at least partly. Before 31 August, most stores had no first-party data on AI Overviews or AI Mode at all. Now every Search Console property has impressions by page. That turns AI search from a topic of speculation into something you can track monthly.
• Impressions without clicks are an incomplete picture. Because AI-specific clicks are not separated, you need to compare AI impressions with Web clicks for the same pages and dates, and with your own analytics, to judge whether visibility turns into visits and sales.
• Collection and guide pages may matter more than product pages. AI answers to shopping questions often summarise a category or a buying decision. Pages that explain a category clearly, such as a well-written collection page or a buying guide, are natural candidates to be cited. Check which of your pages actually appear before deciding.
• Product data still feeds shopping results. For product listings in Google's shopping surfaces, Merchant Center feeds matter as much as page content. Our earlier post on AI Mode shopping for holiday 2026 covers feeds and conversational attributes.
• Model churn is real. With a new Flash model every few weeks, avoid rebuilding pages around how one model answered one query. Build for accuracy and clarity, which every model rewards.

A monthly AI visibility review for a Shopify store

Here is a routine that takes about an hour a month and uses only Search Console and Shopify.

1. 1Open the generative AI performance report for Search and set the date range to the last 28 days, compared with the previous period.
2. 2Switch to the Pages view and export the table. Group URLs by type: collections (/collections/), products (/products/), blog posts (/blogs/) and pages (/pages/).
3. 3For your top 20 pages by AI impressions, open the main Performance report, filter to the same page and dates, and note Web impressions and clicks.
4. 4In Shopify Analytics, check landing sessions and conversion rate for the same pages and dates. A page with rising AI impressions, flat clicks and falling sessions is a page being summarised but not visited.
5. 5For pages that are summarised but not visited, ask whether the page offers anything beyond the summary: prices, availability, size guidance, photographs, comparison, a reason to click.
6. 6For important collections with no AI impressions, check the basics: is the page indexed, does it allow snippets, does it have enough descriptive text to be quotable, and are the first products in stock?
7. 7Note the date and the change you make, so next month's review can tell what worked.

Making collection pages worth citing and worth visiting

A collection page has two jobs in an AI-search world. It needs enough clear, accurate text for a model to quote, and it needs to reward the visitor who clicks through with something a summary cannot give them. Those jobs pull in slightly different directions, and the best collection pages do both.

• A short, specific introduction: what is in this collection, who it is for, what distinguishes it, in two or three sentences of plain language. Not a block of keywords.
• Facts a model can quote accurately: price ranges, materials, size ranges, delivery regions, return terms. Keep them consistent with your product pages and structured data.
• A first screen of products that are in stock and selling, so the visitor who clicks from an AI answer sees good options immediately.
• Filters that work on mobile, so a visitor can narrow by size, price or colour without starting again.
• Links to buying guides or size guides that answer the questions the AI answer probably raised.

The third point is where merchandising and search meet. A visitor arriving from an AI answer has often already decided what kind of product they want. If the first row of the collection is sold out or irrelevant, that decision goes elsewhere. SellRank, our Shopify app, ranks collection products by measured performance, such as views, add-to-carts, orders, revenue and conversion rate, with conditions such as in-stock only, and writes the order into Shopify's own sort, so the page a visitor lands on shows products that are actually selling.

Four common misreadings of AI search data

The new reports are useful, and they are also easy to misread. These are the mistakes we expect to see most often as teams start looking at them.

• Treating AI impressions as visits. An impression means a link to your site was shown in an AI feature. It does not mean anyone saw it, read it or clicked it. Judge success by sessions and sales, not by impressions alone.
• Comparing AI impressions with Web impressions as if they were the same thing. One AI answer can show several of your links and still count once. A results page counts differently. Compare each with itself over time.
• Reading a September change as a trend. A new model arrived on 2 September, a citation bug came and went within days, and a spam update ran from 24 September to 8 October. A change in that period may have several causes.
• Assuming the AI contribution pilot applies to you. Unless Google has contacted you about it, it does not. There is nothing to configure or optimise for it today.

The safest approach is to treat the generative AI report as one more lens on the same question you already ask of Search Console: which pages are found, which are visited, and which turn visits into orders. The answer to the last part still lives in your store analytics.

What we are watching next

• Whether Google makes Gemini 3.8 Flash, or a later model, the default for free AI Mode users, which would change how most shoppers experience it.
• Whether Search Console adds AI-specific click data. Search Engine Journal noted regulatory pressure that may require Google to provide it; we will wait for Google's own announcement.
• Whether the AI contribution pilot expands beyond publishers, and whether Google publishes how contribution is measured.
• Whether Google updates its AI features documentation, which has not changed since December 2025, to reflect the new reports and controls.

We will update this post as Google publishes changes. Until then, the useful work is the same as it has been: measurable, accurate, crawlable pages, and a first screen that converts the visitors you do get.

Get help reading your AI visibility

If you want help setting up an AI visibility review for your store, or connecting Search Console data to what actually sells, tell us about it on our contact page. If your collection pages are being found but not converting, try SellRank's free plan on one collection and see the difference on your own store.