What Changed in Digital Marketing: AI Mode Gets Direct Offers, Discover Plans an Overhaul

Key Takeaways

The September 2026 spam update, which began 24 September, finished rolling out on 8 October after roughly two weeks. If your traffic changed during that window, you can now draw conclusions.
Google is testing a Direct Offer asset inside AI Mode: advertisers supply promo codes by CSV, and Google’s AI decides when and how to surface them as a “Claim one-time code” option under sponsored results. It only runs on campaigns using AI Max or text customisation.
Self-serve Conversion Lift measurement is now available for Search and Performance Max campaigns without needing a Google Ads representative. The thresholds are 1,000 observed conversions and a minimum $5,000 USD campaign budget.
Google’s job postings reveal plans to rebuild Discover from a recommendation feed into a proactive AI concierge, tightly integrated with YouTube, Shopping, Chrome and Android. No launch date is set.
The thread connecting everything this week: Google is building infrastructure for users to act, buy, and decide within its own surfaces rather than clicking through to yours.

Four things happened in the past week worth knowing about. None are dramatic on their own. Together they continue the pattern that has been running through every Google update in 2026: more of the commercial and transactional layer is moving inside Google.

The September Spam Update Is Done

On 8 October, Search Engine Roundtable confirmed the September 2026 spam update had finished rolling out. It began 24 September and took roughly two weeks, making it one of the longer spam update rollouts of the year. The previous spam updates in 2026 wrapped up in about two days.

We covered what to check when it launched in the earlier October roundup. If you followed the advice there and waited two weeks before drawing conclusions, the window is now closed. Traffic changes that settled before 8 October reflect the update’s final state.

Who took the hit

The longer rollout and the SAFE system Google deployed alongside this update suggest it was a more complex evaluation than earlier 2026 spam updates. SAFE uses a language model to assess whether content violates the intent of Google’s policies, not just the letter of specific rules. Sites running high volumes of AI-generated content without genuine editorial judgment applied to it were the most exposed.

If your traffic held through 8 October and the pages in question are not in Manual Actions, there is nothing to change. If you saw a measurable drop and it has persisted past the rollout window, start in Google Search Console’s Pages report, not your content editor.

Self-Serve Conversion Lift: The Rep Requirement Is Gone

On 7 October, Search Engine Land reported that Google had rewritten its Conversion Lift help page to include Search and Performance Max campaigns and dropped the requirement for a Google representative.

Previously, getting Conversion Lift on Search or PMax required going through a rep. For smaller and mid-size accounts without dedicated account managers, that was a hard stop.

The eligibility requirements are now:

At least 1,000 conversions observed by the account (supplementary data does not count)
A minimum campaign budget of $5,000 USD
At least one compatible conversion action
The campaign cannot be in another lift study at the same time (Brand Lift, Search Lift, or Conversion Lift simultaneously is not allowed)
iOS-targeted App campaigns and Travel Ads are not supported

What a Conversion Lift study actually tells you

Most Google Ads accounts optimise toward conversions tracked in the platform. The problem is that those numbers include people who would have converted anyway, without ever seeing your ad. Attribution models try to account for this, but all attribution models contain assumptions.

A Conversion Lift study runs a holdout test. A portion of your audience is excluded from seeing your ads during the study period. At the end, you compare conversion rates between the exposed group and the holdout. The difference is the incremental lift: the conversions your ads actually caused, not just the ones that happened while they were running.

For Search campaigns, this is especially relevant. Someone searching for your brand name or service already has intent. The question is whether your ad added anything to the conversion, or whether they would have found you through an organic listing regardless.

The practical arithmetic: if an account runs $10,000 a month on Search with a tracked cost per conversion of $40, it appears to generate 250 conversions. If a Conversion Lift study shows incremental lift of 60%, that is 150 incremental conversions and 100 that would have happened without the campaign. The true incremental cost per conversion is closer to $67. That is a meaningfully different number for planning.

Running the study does not require any campaign changes. Set it up once the account clears the thresholds, run it for the recommended duration, and use the result as a calibration on the numbers you report.

Google Is Testing Direct Offers in AI Mode

On 7 October, Search Engine Land reported that Google is testing a new asset type called Direct Offer inside AI Mode. It was in beta as of the report.

The format works like this. An advertiser sets up a Direct Offer asset at the account or campaign level, alongside the existing Promotions, Messages and Lead Forms options. They supply an internal name, a final URL, and a short description of the offer. They can attach multiple incentives, including discounts, loyalty benefits or product bundles, and upload single-use promo codes by CSV. Google’s AI then decides when a particular offer is relevant to a query and writes the shopper-facing copy itself.

When the AI serves the offer in AI Mode, it appears under a sponsored result as a “Claim one-time code” option. Advertisers can set caps by total value or number of claims.

One restriction that narrows the immediate audience: Direct Offers only run on campaigns that use AI Max or text customisation. If neither is enabled on a campaign, the asset does not run.

What the format is designed for

It is designed for lower-funnel commercial intent. A query in AI Mode like “best deal on home security installation” carries clear purchase intent. Rather than sending the user to a standard ad, Google is testing whether surfacing a specific, claimable offer at that moment improves conversion.

The single-use promo code mechanic matters. Codes uploaded by CSV mean the advertiser controls the offer’s reach and total redemption value precisely. It is not a blanket discount visible to everyone who visits the landing page; it is a targeted incentive for users who found the ad through that specific AI Mode placement.

For New Zealand advertisers, the caveat is the same as every AI Mode format so far: US beta, no confirmed international rollout date. What is worth doing now is deciding whether your offer structure suits it. If your business runs time-limited promotions or discount codes already, you have an asset type ready to test when it becomes available here.

One thing to watch as the format scales: Google’s AI writes the user-facing copy from the advertiser’s description. The advertiser provides intent; the model provides execution. If the model writes copy that misrepresents the offer, the compliance question sits with the advertiser. The Advertising Standards Authority in New Zealand treats inaccurate advertising as the advertiser’s responsibility regardless of which tool generated the copy.

Google Is Rebuilding Discover as an AI Concierge

On 12 October, Search Engine Journal reported that Google’s job postings describe plans to rebuild Discover into what Google calls a “proactive, hyper-personalised concierge experience that helps users make progress on their goals.”

The current Discover feed is a recommendation engine. It surfaces articles, videos and content based on topics you have signalled interest in through search and browsing history. The concierge model is different in kind, not just degree. Instead of recommending content to browse, it would anticipate needs, track ongoing goals, and surface information, products, services and next steps, drawing on a unified Google ecosystem including YouTube, Shopping, Chrome and Android. The postings explicitly target younger users, which signals a competitive framing against TikTok and Instagram Reels.

Google Discover already serves a large mobile audience, and publishers and content marketers depend on it as a meaningful traffic source.

What the traffic question looks like

The concern is not new, but the architecture being described is more explicit than earlier Discover changes. An AI concierge that answers questions inside Discover, recommends products from Shopping and content from YouTube, has less reason to send a click to an external article or blog post than the current feed does.

Google has included language about “elevating human creators” in the posting, which tends to appear when the company is managing the message around a feature that reduces external clicks. The core tension is the same one AI Overviews already raised: a system designed to help users reach their goals inside Google’s own surfaces does that most efficiently by keeping them there.

No launch date is attached to any of this. Job postings describe intent, not shipping schedules. The Discover concierge could be two years away. The direction, however, is clear.

What content marketers should do now

Nothing reactive. A Discover overhaul signalled by a job posting is not a reason to rebuild your content strategy this month.

What is worth doing is establishing a baseline. Open GA4, go to Acquisition, then Traffic Acquisition, and check what Discover is currently contributing to your organic sessions. If it is sending 20% of your sessions, you have material exposure. If it is sending 2%, you have less to worry about.

The longer-term implication is the same one that applies to AI Overviews and AI Mode: content that helps a person take a specific next step is harder for an AI concierge to replace than content that only describes a topic. A proactive AI system still needs source material to be proactive about. Being the clearest, most specific answer in your subject area remains the strongest structural position, regardless of which Google surface changes around it.

What Connects All Four

The pattern this week is the same one running through Google’s second half of 2026.

Google is not just helping users find things. It is building infrastructure for users to do things inside Google: claim a promo code, measure the actual lift from an ad, track a goal, watch a related video, buy a product. Direct Offers in AI Mode, self-serve incrementality testing, and a Discover feed that anticipates needs rather than passively recommending content are all parts of the same architecture.

For advertisers, the near-term picture is mostly positive. More surface area for commercial intent inside Google means more places to connect with a buyer who has already decided. The Direct Offer format, when it reaches New Zealand, is particularly well suited to service businesses that already run seasonal promotions or referral codes.

For publishers and content marketers, the calculus is different. Every time Google builds a layer that satisfies a user’s intent inside its own surface, click-through rates on content that used to satisfy that intent decline. That has been true of social feeds, AI Overviews and Featured Snippets before them. The Discover concierge is the next chapter of the same story, not a new one.

What To Do This Week

Close out your September spam update review. The update finished 8 October. Changes that happened between 24 September and 8 October are now the update’s final verdict, not mid-rollout noise. If your traffic is down on pages that were previously stable, start with Manual Actions, then the Pages report in Search Console.

If your account qualifies, set up a Conversion Lift study. The thresholds are 1,000 observed conversions and a $5,000 USD campaign budget. If you clear those numbers on a Search or PMax campaign, you can now run a holdout test without needing a representative. The result tells you what your spend is actually causing versus what it is correlating with.

Familiarise yourself with the Direct Offer asset format. It is a US beta and is not available here yet. If your business already runs discount codes, referral codes or time-limited promotions, the format is a natural fit when it opens up internationally.

Check your current Discover traffic in GA4. Note the volume now. If the Discover concierge rolls out and changes traffic patterns in 12 to 18 months, you will want a baseline to compare against.

We watch these updates each week and apply them directly to the accounts we manage. If you would rather someone else ran that process on yours, get in touch.

Frequently Asked Questions

Is the September 2026 spam update still running?

No. It finished rolling out on 8 October 2026, roughly two weeks after it began on 24 September. Any rankings changes that occurred during that window have settled. If your traffic is different from before 24 September, that is the update’s final verdict.

What is the Direct Offer asset in Google Ads?

A new asset type, currently in beta in the US, that lets advertisers supply promo codes and offer descriptions. Google’s AI selects which offer to show and writes the user-facing copy. In AI Mode, it appears as a “Claim one-time code” option under a sponsored result. It only runs on campaigns using AI Max or text customisation.

How do I qualify for self-serve Conversion Lift in Google Ads?

You need at least 1,000 conversions observed by your account (supplementary data does not count), a minimum campaign budget of $5,000 USD, and at least one compatible conversion action. iOS App campaigns and Travel Ads are not supported. A campaign can only be in one lift study at a time.

What is a Conversion Lift study and why does it matter?

A Conversion Lift study runs a holdout test: a portion of your audience is excluded from seeing your ads, and at the end you compare conversion rates between the exposed group and the holdout. The difference tells you how many conversions your ads actually caused, rather than just correlated with. It is the clearest available measure of incremental return from a campaign.

What is Google Discover’s AI concierge plan?

Based on job postings reported by Search Engine Journal on 12 October, Google is planning to rebuild Discover into a proactive AI assistant that helps users work toward goals, drawing on YouTube, Shopping, Chrome and Android. No launch date has been set. It is a direction signal from hiring intent, not an announced product.