Q3 2026 Platform Deadlines Every Marketer Must Track

Six Q3 2026 platform deadlines are circulating in marketing teams right now, and they are not backed by the same kind of evidence. Three sit on a published vendor document that states the exact date. The other three thin out fast: an email Google sent to users with no page behind it, a vendor blog post that commits to a month but never a day, and one date nobody could trace to a vendor at all. Sorting them by confirmation status rather than by date is what turns the list into something a team can act on.

That distinction is the whole point of this page. A screenshot of a practitioner post and a Google help article both read like facts in a Slack thread, and both get pasted into the same planning doc. They are not the same input. Building a sprint around a date that no vendor has published is how a team spends two weeks preparing for a change that lands somewhere else, or on a different feature entirely.

What follows is one dated calendar for the changes announced between now and the end of Q3, each carrying its confirmation status, what breaks if it is ignored, who on a marketing team should own it, and a link to the vendor page or the deep-dive. Where an item already has a full write-up on this site, it gets one line here and a link, not a rewrite. Where an item could not be verified, this page says so plainly instead of dressing the gap up as a hedge.

Most Q3 roundups sort by date, which quietly implies that every row carries the same weight. That is the assumption that costs a team a sprint. These six items sit in four genuinely different evidence classes, and the right operational response follows the class, not how soon the date arrives.

Five of the six items get a row. Dates are as announced by the vendor or, where labelled, as reported. Weekdays are on the rows because two of the five land on a Monday or a weekend, which decides who is actually at a desk when the change hits.

The sixth item is missing on purpose. The August 24 Amazon Business Solutions Agreement transfer and pledging clause is covered in section 07, but it does not get a dated row, because a dated row in a calendar is an assertion and we could not find an Amazon source that supports one.

Those three are the only items here where both ends of the notice period are firmly published, which is why the other three are absent from the comparison rather than estimated. The Google Assistant email has no stated send date — it surfaced in reporting on August 4 and 5, which puts the notice at roughly a month, but that is an inference from when journalists saw it, not from Google. The AI Max blog post commits to a month rather than a day, so there is no second endpoint to measure to. And the Amazon item has no first endpoint at all.

August 9 — ChatGPT Atlas stops working. OpenAI’s Help Center article states the date directly and gives the reason: browser-based agentic capabilities are moving into ChatGPT and Codex. The operationally important line is not the date, it is the migration behaviour. Bookmarks do not transfer automatically, and the same applies to open tabs and browsing history — OpenAI tells users to export bookmarks before August 9 and import them into another browser. ChatGPT conversation history is stored separately from Atlas browser data and continues in ChatGPT itself. The replacement path OpenAI names is the ChatGPT desktop app, with the Chrome extension and sidebar where available. Search Engine Land corroborated the date independently, citing an OpenAI statement on X. If you want the context on what Atlas was trying to be before this, we wrote a strategy guide to the browser itself.

August 17 — Google Ads target-based bidding gets stricter for budget-limited campaigns. Google’s help document and its FAQ companion both state the date. The one-line version for a calendar: campaigns with a Limited by budget status running Target CPA, Target ROAS, or Demand Gen Target CPC will deliver more consistently toward the target you have set, including when you adjust budgets — and Google states it will not adjust your targets or budgets on your behalf. Campaigns that are not budget-constrained are explicitly unaffected. That is as much as a hub entry should say. The full mechanics, the campaign-type eligibility table, and what to do with targets that have been sitting untouched for a year are in our full breakdown of the August 17 bidding enforcement.

The Bid Target Adjustment Tool that Google shipped on July 6, 2026 is the tell for how this one is meant to be handled. A vendor that ships a review-and-reset tool six weeks ahead of a change is signalling that the change is expected to move spend for accounts that do nothing. Treat the tool as the deadline, not the date.

This is the entry most Q3 roundups get subtly wrong, so it is the one to slow down on. Google’s Ads & Commerce blog post, updated June 11, 2026, states that campaigns using Automatically Created Assets and the campaign-level broad match setting will continue to be auto-upgraded starting in September 2026. That is a vendor commitment to a month, published on a page you can open. It also states that AI Max will be enabled with settings configured to mirror your legacy setup once the auto-upgrades complete, and frames the release as AI Max moving out of beta.

What Google’s post never states is a day. The widely circulated September 1 date comes from Search Engine Land, on August 6, 2026, which credits the discovery to a practitioner post on X relaying the text of an advertiser-notification email Google sent around August 5. Search Engine Roundtable and PPC Land carry the same claim with the same underlying source. That is a reasonable working assumption, and it is not a vendor-stated date. The practical difference: plan the work to complete before September, and do not tell a client that campaigns flip on the first.

Google’s stated rationale for pushing adoption only means something with its denominator visible. The blog post says AI Max for Search campaigns see an average of 7% more conversions or conversion value at a similar CPA or ROAS when using the full feature suite — search term matching, text customization and final URL expansion — compared to using search term matching alone. That comparison is AI Max against AI Max: full feature suite versus one feature. It is not a comparison against a pre-AI-Max legacy campaign, which is how the number tends to get retold. It is also Google’s own reported figure with no independent audit cited. Useful as a directional signal for how to configure AI Max once you are on it; not a forecast of what the auto-upgrade will do to your account.

If you do not want the upgrade, the lever is upstream of the date: turn off Automatically Created Assets or the campaign-level broad match setting, or turn AI Max on deliberately ahead of the schedule so the transition happens on your terms rather than during a September auto-migration. Either way the audit is the same — list every Search campaign with ACA or campaign-level broad match enabled, and decide per campaign. That audit is the deliverable this month, and it does not depend on whether the day turns out to be the first.

Google is discontinuing Assistant on Android mobile devices, with removal beginning September 4, 2026. The wording is Google’s own, and it is unusually specific about the rollout being gradual rather than instant.

The scope in the same email: once a phone switches to Gemini, the change reaches paired Wear OS watches, Gemini-compatible headphones and earbuds, and Android Auto when it is projected from a mobile device. There is one explicit exception — for cars with Google built-in, Assistant continues to function beyond September 4, 2026. And there is no undo: once availability is removed, the email states you will no longer be able to use or switch back to Assistant on your phone, tablet or paired devices.

Here is the part that matters for how you cite this. At the time of writing, there is no support.google.com help article and no blog.google post carrying this date and scope as an official company document. The strongest primary-strength source is the verbatim text of Google’s own email, reproduced independently by Search Engine Roundtable and 9to5Google, with Search Engine Land and Android Authority carrying the same wording. Multiple outlets quoting identical email text is strong evidence that the text is real. It is still not a page you can send to a client.

One trap to avoid while sourcing this yourself: a support.google.com/assistant/community-guide/ URL surfaces in search for this topic and is marked on the page as community content that may not be verified or up to date. It is forum-authored, not a Google document. Linking it as if it were official is how a Tier B item gets misfiled as Tier A.

Cloudflare’s Content Independence Day post, published July 1, 2026, splits crawlers into three classifications and sets new defaults for each on September 15, 2026. The three classes matter because they replace what used to be a single Block AI bots toggle.

The mechanic that catches people lands on the same date and is easier to skim past. From September 15, crawlers that combine Search with Training are allowed or blocked according to all of their behaviours, enforced by the most restrictive applicable rule — and Cloudflare names Googlebot, Applebot and BingBot as bots that will be blocked for customers who have selected to block Training, either through the new AI-traffic options or through the legacy Block AI bots service. That last clause is the one to read twice. A site that switched on Block AI bots at some point and has not thought about it since is in scope: the setting chosen to keep content out of model training starts keeping out the crawler that feeds classic search. Defensible in some businesses, a catastrophe in others, and either way it should be a decision taken in Security settings before the date rather than a discovery made after it.

On scale, the only figure worth quoting from the post is Cloudflare’s own claim that more than 20% of web domains sit behind Cloudflare — a network-reach number about domains, not a statistic about crawler behaviour or traffic mix. Some secondary write-ups attach a crawler-traffic percentage to this announcement; it is not in the Cloudflare post, so it is not here either. The post also introduces a use= field extending the robots.txt Content Signals vocabulary, with use=immediate, use=reference and use=full to express reuse preference. If AI-crawler posture is now part of your visibility strategy, that pairs with Cloudflare’s AEO visibility and agent-readiness tooling and with how we approach agentic SEO.

A clause said to take effect on August 24, 2026 is circulating among Amazon sellers: an update to the Business Solutions Agreement, reportedly dated May 29, 2026, that would prohibit sellers from transferring their rights or obligations under the BSA, or pledging them as collateral. If real, it would matter — it touches account sales, aggregator deals and revenue-based lending, which is exactly why it spreads.

We could not confirm it on anything Amazon published. What was checked, at the time of writing: the Business Solutions Agreement document itself; Seller Central announcement threads, where the only dated BSA-update thread found discusses a March 4, 2026 update, a different and earlier change; Amazon’s seller announcements blog; and two seller-law-firm blogs, one of which also covers only the March 4 update. Every secondary source repeating the August 24 claim traces back to the same handful of seller-advisory posts, none of which links the operative BSA text or a Seller Central thread.

The real story this quarter is not any single date. It is that vendors have stopped announcing through one channel. Of these six items, one came through a help centre, two through company blogs, one through a direct email with no accompanying public page, one through a help doc with no publish date, and one through nothing traceable at all. A marketing team that only watches help centres will miss the Assistant date entirely; a team that only watches trade press will get the AI Max day and miss that Google never committed to it.

The forward-looking read is that this gets harder rather than easier. One of the six changes here is about who gets to read your content and on what terms; another is about an ad platform moving campaigns onto an AI system by default. Both categories tend to be announced by blog post rather than by product documentation, and blog posts get updated in place — the Google post that anchors the AI Max entry was published in April and updated in June, which is how the February 2027 DSA extension ended up living in the same URL as a September 2026 date. Any Q4 calendar built the same way will need re-verification, not just a new row, because the source itself can change under a stable link.

The practical habit that survives this: whenever you write a date into a plan, write the URL next to it and the tier next to that. When someone challenges it in three weeks, you re-open one link instead of re-running the whole investigation. That is the same discipline we apply to the AI model-deprecation side of this same calendar problem, where sunset dates move and the only defence is a sourced list. For the paid-media rows specifically, it sits alongside two other Google Ads changes worth tracking this month — the corporate-email requirement affecting agency account access and the third-party pre-screen arriving for the Search Partner Network. If you would rather someone else held this list, that is part of how we run paid media for clients.