What's new in Google Analytics 4 (2026): the 7 changes that actually matter

Google has shipped more changes to Google Analytics 4 in the last 18 months than in the entire run of Universal Analytics before it. Most of those changes were tiny tweaks that don't affect anyone. A few of them genuinely change how you should be using the tool, and one of them probably broke your remarketing without you noticing. Here's the plain-English roundup of the seven changes that actually matter for small businesses and the agencies who help them — what each one means, who it affects, and what you need to do this week.

1. Universal Analytics is officially, irreversibly gone

Universal Analytics stopped collecting new data in July 2023, but the historical data stayed accessible inside the old UA properties for another year. That ran out on July 1, 2024, when Google deleted all UA data permanently. There is no recovery option, no archive, no "please bring it back." If you didn't export your UA historical data to BigQuery, CSV, or a third-party warehouse before that date, those years of behavioural history are gone for good.

What this means in early 2026: most businesses still doing year-over-year comparisons against the UA era are comparing apples to oranges. GA4 measures sessions, users, and conversions with different definitions than UA did. If your weekly report still references "sessions vs last year" using exported UA numbers, you're comparing two different yardsticks — and the comparison is almost always misleading. The honest baseline year is now whichever year you got GA4 running properly. For most small businesses that's 2024 or 2025.

Action item: stop comparing your 2026 GA4 numbers against your UA numbers. Either pick GA4's first full year as the baseline, or work in shorter (90-day, 180-day) rolling comparisons until you accumulate enough comparable GA4 history. Your future-self will trust the data more if you stop trying to bridge the old and new world.

2. "Conversions" got renamed to "Key Events" (and the meaning changed)

In late 2024 Google renamed every "conversion" in GA4 to "key event," and the distinction is not just cosmetic. A "key event" in GA4 is now any event you've marked as business-important — purchases, form submits, scroll-depth milestones, anything. The word "conversion" is now reserved for events imported into Google Ads, where it has a stricter advertising-specific meaning. The two terms are no longer synonymous.

What this means: every old tutorial you find online that says "set up a conversion in GA4" is using vocabulary that no longer matches the interface. The clicks are the same — Admin → Events → toggle on "Mark as key event" — but everything is labelled differently and the columns in your reports moved. Tutorials that haven't been updated since 2024 will leave you hunting for buttons that have been renamed.

Action item: open Admin → Events in your GA4 property and confirm your business-critical events (purchase, generate_lead, contact, signup) are marked as key events. The toggle is now called "Mark as key event" — if you don't see your critical event in the list, you may need to create it as a custom event first.

3. Google Consent Mode v2 is mandatory for EU advertisers (and almost certainly affecting your tag coverage)

Consent Mode v2 became required in March 2024 for any advertiser using Google Ads remarketing or measurement features on EU/EEA traffic. The short version: if a visitor declines analytics cookies, your GA4 tag must either suppress the request entirely OR send a cookieless "ping" with stripped identifiers. The version-2 update introduced two new consent signals (ad_user_data and ad_personalization) that are now required separately from analytics_storage.

What this means in practice: if your cookie banner hasn't been updated since 2023, your remarketing audiences in Google Ads have likely been quietly shrinking because Google is rejecting non-compliant data. We've seen Google Ads audiences drop 40-60% in the months after Consent Mode v2 enforcement kicked in for sites that didn't update their banner. The traffic didn't change — Google's willingness to accept the data did.

Action item: ask your cookie-consent vendor (Iubenda, Termly, CookieYes, OneTrust, Cookiebot) whether your banner is sending all four Consent Mode v2 signals — analytics_storage, ad_storage, ad_user_data, ad_personalization. If the answer isn't an immediate yes, update the banner today. The fix is usually a single toggle in your consent tool. This is the most common silent data problem we see in 2026.

4. AI-generated insights now appear in every GA4 property

Google has rolled out an automated "Insights" panel in every GA4 property, where their own AI surfaces anomalies and trends without you asking. It lives on the Home tab and updates daily, calling out things like "Organic Search traffic increased 38% week-over-week" or "Pageviews for /pricing dropped sharply on Tuesday." The feature is free and on by default.

What this means: Google has finally admitted that 95% of GA4 users don't actually use GA4 — they just want a Monday-morning summary. The Insights panel is their first real attempt to deliver that, and for the basic "what changed this week" question, it's surprisingly useful. The limitations: the panel doesn't explain WHY anything moved, the language is sterile, and the anomalies don't link to specific suggested actions. You still have to interpret the changes yourself. (This is also exactly the gap Plainly was built to fill — we connect to the same data, but we write the WHY in plain English and tell you what to try next, automatically by email every Monday.)

Action item: log into GA4, click Home, and see if there's an Insights panel on the right side of your dashboard. If yes, glance at it daily for two weeks. You'll quickly learn whether Google's own auto-detection is enough for you, or whether you need a layer that translates the anomalies into actions.

5. The Reports library lets you finally customise the navigation

One of the most common complaints about GA4 since launch was that the left-hand navigation was bloated and impossible to clean up — every property looked the same and most of the menu items were irrelevant for any given business. Google quietly fixed this in 2024 by shipping the Reports → Library editor: a way for any property editor or owner to add, remove, hide, and reorder every report in the navigation.

What this means: you can finally hide "User attributes," "Tech details," "App developer" and all the other reports your team will never open, and pin the 3-4 reports you actually use to the top. The five reports we recommend pinning for a small business: Traffic acquisition, Landing page, Pages and screens, Events, and Key events. Everything else can go in a "More" collection or be hidden entirely.

Action item: in GA4, click Reports → Library (bottom of the left nav) → Edit the default Collection. Move your top reports to the top, hide everything you don't use. It's a 10-minute job and it makes every subsequent GA4 visit dramatically less overwhelming.

6. Custom channel groups are now widely available

GA4 originally forced everyone to use Google's standard channel definitions — "Organic Social," "Paid Search," "Direct," etc. — with no way to override them. That changed in 2024 when Google rolled out custom channel groups to all standard (free) properties. You can now define your own rules for which source/medium combinations roll up into which channel name.

What this means: if you've ever been frustrated that traffic from Substack shows up as "Referral" instead of "Email," or that your specific newsletter campaigns get lumped in with random press mentions, you can now fix it. Admin → Channel groups → Create new channel group lets you define rules like "if source contains 'substack' OR medium = 'newsletter', call it Email." The custom group lives alongside the default one, so your historical reporting stays consistent.

Action item: only worth doing if your reports are actually misleading you because of mis-grouped traffic. If 90% of your traffic falls into the standard channels correctly, leave it alone. If you're constantly explaining away the same mis-classifications to clients or your team, build a custom channel group and switch the reports to use it.

7. Attribution settings changed — you might be measuring two different things

In 2024 Google rolled out a major change to default attribution models in GA4: "Data-driven attribution" became the default for new properties, replacing "Last click" which was the silent default for years. This change applies to acquisition reports, conversion reports, and what gets pushed back to Google Ads. The catch: properties created before the change kept their old setting unless someone manually updated it.

What this means: if you compare attribution numbers between two GA4 properties — say, yours and a client's — they may be using different attribution models without anyone realising it. "Data-driven" tends to credit upper-funnel channels (Social, Display) more generously than "Last click" does, so the same exact customer journey can look different depending on which model is active. This causes endless confusion when an agency reports different conversion numbers than the brand's own GA4 dashboard.

Action item: open Admin → Attribution settings and check which model is set. For most small businesses we recommend explicitly choosing "Data-driven" (it's free and more accurate) but knowing which one you've picked. If you switch, expect a 10-30% re-distribution of credit across channels — that's the model adjusting, not your actual marketing changing.

Every quarter, Google ships changes that subtly redefine what your numbers mean. If you don't audit the settings annually, you're slowly drifting into reports that measure something different than they did last year.

What didn't make the list (but might still matter to you)

A handful of smaller changes shipped that we deliberately left off the headline list because they only matter to a subset of users:

• Increased free BigQuery export quota. Google raised the free daily export limit for standard GA4 properties — useful only if you're already piping GA4 into BigQuery for custom reporting. If you've never opened BigQuery, ignore.
• Subproperties and roll-up properties became more accessible. Useful for agencies and multi-brand businesses, still gated behind GA4 360 ($50k+/year). Not relevant for solo operators.
• Cross-channel data-driven attribution lookback window changed. Default lookback for paid conversions adjusted from 30 days to 90 days. Mostly affects ad-heavy businesses comparing year-over-year ROAS.
• New 'Modeled conversions' transparency. GA4 now flags which conversions are observed vs. modelled. Good if you've ever wondered why GA4 numbers seem 'too clean' — modelled conversions are an estimate, not a count.

The 20-minute 2026 audit checklist

If you read nothing else above, run through this checklist once and you'll be ahead of 90% of small businesses on GA4 hygiene in 2026:

• 1. Confirm Consent Mode v2 is sending all 4 signals. Most common silent data problem this year. Ask your cookie-consent vendor.
• 2. Check Admin → Attribution settings. Pick 'Data-driven' explicitly. Document the choice.
• 3. Open Admin → Events. Mark your business-critical events as Key events.
• 4. Trim the Reports library. Pin the 5 reports you actually use, hide the rest.
• 5. Confirm internal-traffic filter is active. Filter out your own visits before they pollute the data.
• 6. Time zone matches your business. If you compare GA4 to Shopify or Stripe, all three must agree on what 'a day' is.
• 7. Glance at the Insights panel weekly. Free anomaly detection; decide whether it's enough for you or if you need a deeper layer.

Two of these (Consent Mode v2 and attribution settings) are the most-overlooked problems we see in 2026 audits, and both can quietly distort your numbers by 20-40% without anyone noticing. Fixing them takes 20 minutes combined and pays for itself the first time you make a decision based on cleaner data.

The bigger trend: GA4 is finally maturing

Three years after Universal Analytics shut down, GA4 is finally maturing into something a small business can actually use without an analyst. The Insights panel, the Reports library, custom channel groups, and clearer attribution settings collectively close most of the "GA4 is broken" complaints from 2022-2023. The remaining gap — translating the data into plain-English narrative and recommendations — is what tools like Plainly handle on top.

Looking ahead: in 2026 Google is heavily betting on AI-assisted analytics. Expect the Insights panel to get smarter, expect more LLM-generated summaries inside GA4 itself, and expect the boundary between "raw analytics" and "AI-summarised insights" to keep blurring. The smart move for small businesses is to spend less time learning GA4's interface and more time defining which decisions you actually want analytics to answer — then pick whichever tool (GA4's Insights, Plainly, a consultant) gets you those answers in the format that fits your weekly workflow.

Or, just let Plainly summarise the changes for you each week

Plainly reads your GA4 data every Monday and writes you a plain-English summary of what changed week-over-week, what likely caused it, and what to try next — using the exact channel and page-level breakdowns covered above. No dashboards to learn, no settings to babysit, no '7 changes that actually matter' to memorise. We do the interpreting; you make the decisions. Try the free demo from the homepage — no credit card required.