The March 2024 core update isn't a penalty on affiliate sites for being affiliate sites. It goes after pages that could have been written without anyone opening the product. If you run a review site, that's the test to hold every page against this week. Could someone who never logged in have written it? If the answer is yes, the page is exposed, and it makes no difference whether a person or a language model typed it.
Google has been plain on that last point. Its new scaled content abuse policy covers mass-produced pages however they're made. So if your plan is to replace AI drafts with cheap freelance drafts, you'll pay for the same problem twice. The affiliate sites that come through this will have cut their page count, put a named person who used the software behind every review, and shown testing evidence a competitor's writer can't copy.
It's Friday 15 March 2024 and the rollout is still running. Nobody knows the final impact yet, us included. What follows covers what Google announced, the three new spam policies and their dates, why the standard software roundup matches the pattern Google describes, what real testing evidence looks like on a software review, and how to make author bylines count for something. The second half covers pruning, choosing which products to promote, and building traffic that doesn't depend on Google.

Google's 5 March announcement is aimed squarely at pages nobody tested
On 5 March 2024 Google announced two things together, a core update and a set of new spam policies. Google described the core update as more complex than usual, with changes to several core ranking systems, not one. The biggest structural change is that the helpful content signals no longer sit in a separate system. They're now part of core ranking.
That matters for march 2024 core update affiliate sites in particular. Under the old setup, a site-wide helpful content classifier could hold a domain down until it improved. Now the same idea runs through the main ranking systems, which Google says use a range of signals to judge whether a page is helpful. You won't find a single switch to flip back. Unhelpful pages lose ground inside the system that ranks everything else.
Google said the rollout could take up to a month. It also said it expects the combined changes to cut low-quality, unoriginal content in search results by 40%. That figure is Google's own estimate of the effect on results. It's not a promise about any single site, and it isn't a measured outcome yet.
So is your site being hit by the core update or by the spam policies? Without a manual action you can't be sure, and during a month-long rollout nobody can. What you can do is check your pages against what both target. They target the same thing, which makes the job simpler than it sounds.
Three new spam policies, and only one of them gives you a grace period
Google added three policies to its spam rules. Two are being enforced now. One, site reputation abuse, starts on 5 May 2024, which gives you about seven weeks to check where your content sits. Here's each one in plain terms.
- Scaled content abuse. Producing many pages mainly to rank in search rather than to help people, whatever method produced them. Enforcement started straight away.
- Site reputation abuse. Third-party pages hosted on a strong domain, with little oversight from the host, published to borrow that domain's ranking power. Enforcement starts on 5 May 2024.
- Expired domain abuse. Buying a lapsed domain and filling it with low-value content to rank on the old site's reputation. Enforcement started straight away.
Bought an expired domain for your review site? The policy targets the intent and the content, not the purchase. Buying a lapsed domain and building a real, tested review site on it is a different thing from buying an old charity's domain and filling it with casino reviews. The question is whether your content fits what the domain used to be and stands up on its own. If the old backlinks are the only reason the site ranks, you're exposed.
Scaled content abuse covers human writers too
Does Google now penalise AI-written content? No, and the policy goes further than that. Google's wording is about the purpose and the scale of the pages, not the tool that made them. Pages made mainly to rank, at volume, and with little value to the reader are in scope whether a model, an offshore content team or a template with swapped keywords produced them.
Take a site with a hundred pages titled "best email tool for dentists", "best email tool for plumbers", "best email tool for yoga studios" and so on. Each page lists the same products with a new opening paragraph. Freelancers wrote every one by hand. That site carries the same risk as one that generated them in an afternoon. Changing the writer doesn't change the pattern.
Site reputation abuse matters if you publish on someone else's domain
This one's for creators who place reviews, coupon pages or "best of" lists on a publisher's subdomain, a news site's partner section or a host site that rents out space. If the host site has little involvement in the content and the pages are there to ride its authority, that's the pattern Google names. Google will enforce it from 5 May 2024.
Before then, list every page you have on a domain you don't own. For each one, ask three questions. Does the host edit or oversee it? Does it fit what the host site is about? Would it rank on its own merits? If the honest answer to all three is no, plan to move it to your own site or take it down before May. Don't wait to see what happens.
Republishing your own work elsewhere is a different case. Content syndication with clear ownership, a canonical link back to the original and a host that chose to run it is ordinary publishing. Renting a strong domain's authority for pages the host never really looked at is the thing the policy describes.
Why the typical software roundup looks like scaled content to Google
Open five "best email marketing software" pages and you'll see the same template. The same ten tools appear in roughly the same order. Feature lists are rewritten from the vendors' own pages. Screenshots are the vendor's marketing images, or nothing at all. Every item has a "best for" tag, and every tool is somehow best for something. The top carries a "last updated" date, and nothing on the page changed on that date.
Then the site repeats the page across hundreds of URLs. There's one for every pairing of two tools, one for every industry and one for every "alternative to" query. Swap the product names on two of those pages and you can't tell them apart.
Both halves of the 5 March announcement catch this. The helpfulness signals now inside core ranking ask whether a page adds anything beyond what already exists. A rewritten feature list adds nothing, because the vendor's page already says it better. The scaled content policy asks whether pages were made at volume mainly to rank. Hundreds of near-identical comparison URLs answer that question for you.
None of this means comparison pages are banned. A single comparison of two tools you've used side by side, with your own results, is useful to a reader. The problem is the template replacing the testing. Once you spot that in your own archive, you know which pages need work.
First-hand testing is the one asset a scaled site cannot copy
A competitor can copy your structure, your headings and your shortlist in an hour. They can't copy the afternoon you spent inside the product. That's the one thing that sets a tested review apart from a scaled one, so make it visible on the page.
What counts as evidence on a software review is fairly specific. These all come from use, and none of them can be rewritten from a vendor's site.
- Screenshots from your own account, with your own test data in them, not the vendor's demo images.
- Setup times you measured, such as how long from signup to a sent first campaign.
- Support tickets you raised, how long the reply took and whether it solved the problem.
- What broke, what confused you and what you had to work around.
- What the product couldn't do, stated plainly.
- Data from your own test sends or campaigns, such as inbox placement on your test addresses or real open and click counts.
What testing a white-label email platform actually involves
Here's what a proper test of SendTo looks like, since we know our own product. You'd sign up as a partner and set up the branded portal under a test brand. Then you'd build a campaign from a brief and see what the HTML email looks like, and how the one-click remix changes it. You'd run the AI reviewer and note what it flagged in your subject line, links and structure before launch.
Next comes the list. Import a messy spreadsheet and see whether the columns map themselves. Interrupt the upload and check whether it resumes. Count how many addresses verification removes. Then send to seed addresses you control at different mailbox providers and record where each one lands. Note how sending is spread across brand-matched domains, and how warmup applies to a new sending domain. Finally, open the reports. Compare the human opens shown against what you saw yourself, since the reports exclude security scanner opens. Our SendTo review shows how much detail this gives you to write about.
Write down the limits as well. SendTo is built for agencies selling to clients under their own brand. A solo creator who wants a simple personal newsletter is not who it's for, and a review that says so is more useful than one that doesn't. If the reader asks about cost, send them to the pricing page instead of quoting a figure that will go out of date.
Named authors only help when the name stands for real experience
A byline does nothing by itself. It counts when the name belongs to someone a reader could check, who has used the products they write about. Google's quality guidance has asked for years whether content shows experience and expertise. Google hasn't said how, or whether, any ranking system measures a byline directly, so don't treat an author box as a ranking trick. Treat it as proof for the reader, and as a record that matches what's on the page.
A credible author signal has four parts. It's a real person. Their bio says which tools they've used, in what job, and for how long, for example "ran email for three agency clients on two platforms since 2019". Their byline is consistent across the site, with an author page listing everything they've written. And readers have a way to contact them.
Invented personas are a risk, not a shortcut. A made-up "senior marketing analyst" with a stock photo is easy to expose with a reverse image search, and once a reader finds it, every claim on the site looks false. It also contradicts the whole argument of your page, which is that a real person tested this. If your reviews are written by a team, name the person who did the testing on each one, even if it's you on every page.
Prune and merge before you publish anything new
Cutting pages during an update feels backwards, but it's the most useful thing you can do this month. Every thin page on your site is one more example of the pattern Google described on 5 March. A small archive of tested reviews is much easier to defend than a large one built from templates. Here's the process we'd run during the rollout weeks.
- Step 1: Export every URL. Pull your sitemap and twelve months of Search Console performance data into one spreadsheet, with clicks, impressions and referring domains next to each page.
- Step 2: Group pages by intent. Put every page that answers the same search need into one group. Ten "best email tool for [trade]" pages usually serve a single intent.
- Step 3: Mark each page keep, merge, rewrite or remove. Keep it if it's tested and earns its place. Merge it if another page in the group does the job better. Rewrite it if you've used the product and can add evidence. Remove it if none of that applies.
- Step 4: Redirect merged pages to the strongest one. Use a 301, and move any real testing material across before you switch the old page off.
- Step 5: Noindex or delete the rest. A page with no traffic, no links and no testing behind it has no reason to stay.
Step 2 is where it usually goes wrong. Two pages with different keywords often answer the same question, and two that look alike sometimes don't. If you want a method, our guide to keyword intent mapping groups pages by what the searcher is trying to do, not by the words they type.
So should you delete thin pages or rewrite them? Rewrite only when you can add something that came from using the product. If you've never logged in, a rewrite just gives you a better-written version of the same untested page. Delete it, or merge it into a page you can back up.
Do not read rankings mid-rollout as a verdict
The rollout started on 5 March and Google said it could take up to a month. Rankings will keep moving until it ends. A page that drops on Tuesday and comes back on Thursday hasn't told you anything yet.
Should you make changes before it finishes? Structural ones, yes. Cosmetic ones, no. Keep a dated log of which pages moved, when they moved and what you edited. Fix the structural problems now, meaning clusters of thin pages, untested reviews and content on other people's domains. Don't rewrite headings and intros every day. If you do, you won't know which change caused what once rankings settle.
Recommend fewer products, and pick ones whose commission survives a weaker month of traffic
Going from 300 pages to 60 changes the maths. Each recommendation now carries more of your income. A roundup of ten tools you skimmed spreads the reader's trust thinly. A review of three tools you've run real campaigns on puts all of it in one place.
That's why the commission structure matters more once you've pruned. One-off payouts reward volume, because you need a steady stream of new clicks every month just to stay level. A weaker month of search traffic takes exactly that away. Recurring commission on a product the customer keeps paying for works differently. A referral keeps paying for as long as the customer stays, so a quiet month doesn't take your income back to zero. That isn't guaranteed, since any customer can cancel in any month. But the model fits a site that runs on trust rather than volume.
For a smaller site built on fewer, deeper reviews, three questions decide which programmes to join. Have you used the product enough to write a tested review? Does your audience actually buy it? Do customers stay, so the recurring part is worth something? Our piece on software affiliate programs explains how recurring structures work and what to look for in the terms.
Search traffic is now one channel, not the whole business
A month-long rollout with no known outcome shows the cost of relying on one source of traffic. If one company's ranking systems decide how your month goes, you can't plan your business around it. And this won't be the last update.
An email list is the part you control. A reader who signs up from one of your reviews still hears from you next month, whatever happens in search. Give them a reason to sign up that comes from your testing, such as your testing log template, a monthly note on what changed in the tools you cover, or a checklist for choosing a platform.
Set the list up properly from the first signup. Use signup forms protected against bots, because a visible form on a review site attracts fake signups, and fake addresses damage your sending reputation. Turn on double opt-in so every address has confirmed it wants to hear from you. You'll lose a few people who never click the confirmation link. In return every address on the list is real and has given consent, which matters under GDPR if you have UK or EU readers.

Search itself is changing at the same time. AI-generated answers at the top of results summarise what's already published, so a rewritten feature list gives them nothing new to cite. Your own testing data does. Our article on generative engine optimization explains how getting cited in those answers differs from ranking a page.
Your next step: rebuild around the products you can actually vouch for
The rollout is still running and nobody knows how it will end, us included. You don't need to know. Google has been clear enough about what it's targeting for you to act this week.
- Audit this week. Export every URL and group pages by intent.
- Cut or merge thin pages, and redirect any that have earned links.
- Put a named author behind every review, a real person who used the product.
- Add testing evidence such as screenshots, timings, support tickets and the limits you found.
- Check any hosted or syndicated content on other domains before 5 May 2024.
- Start an email list with bot-protected forms and double opt-in.
If you review marketing or email software, SendTo is a product you can test end to end. You can set up a branded portal, build a campaign from a brief, run the AI reviewer, import a messy list, watch verification and warmup work, and read reports with scanner opens removed. Everything you'd write comes from your own account. If cost comes up, point readers to the pricing page.
Here are the programme terms in full. You earn 20% of the partner subscription each agency you refer pays SendTo, every month for as long as that agency keeps paying. You never earn a share of what the agency's own clients pay the agency. Attribution lasts 90 days from the first click, and the first touch wins. Commission is paid monthly by PayPal.
Write the tested review, then earn 20% of the partner subscription of every agency you refer, every month they pay.
Join the SendTo affiliate programme
