AdSense in 2026 still pays for pages a human can finish. It does not pay for a farm of "14 best AI tools" with the same H2s under six slugs. This is the working procedure we use on this site: what gets a property limited, how the article template is laid out, how to reserve space so units do not shove the text down, and the one-URL audit that decides whether a page keeps its ads or gets unpublished.
This is not Google's policy text. Program policies change and they are the source of truth — read the current version in your account before you ship anything here. What follows is the part policies do not cover: the editorial and layout discipline that keeps a small publisher out of the review queue.
The test that governs everything below: strip every ad unit from the page. If a useful article does not remain, you do not have an inventory problem, you have a page problem. That is the same bar we apply to affiliate placements.
Three things put a small site into "limited ad serving" or a policy action, and none of them are exotic.
The first is thin, duplicated bodies. A 200-word stub with a doorway title, a US listicle rewritten with synonyms, or the same how-to fanned out into five locales as machine translation. The reviewer does not need a tool to see it — the H2s are identical across slugs and the paragraphs say nothing a person could act on.
The second is layout where inventory outweighs content. Ads before the answer, sticky units whose close control is off-screen on a small phone, in-article units dropped mid-sentence, or a "Recommended" stack that sits between every H2. If the density looks like an ad network rather than a blog, that is how it will be read.
The third is invalid traffic. Reviews look at intent, not just clicks: did the page exist to answer a query, or to host units? Buying traffic, asking readers to "support us by clicking", clicking your own units to test placement, or pushing a social burst at a page with no organic demand all land in the same bucket. Test placements with the preview tools, never with real clicks.
There is also a category problem that indie sites walk into by accident. Do not mix adult, copyright-gray, or made-for-AdSense doorway hubs with a SaaS how-to under one property. One site personality. If a slug cannot stand as a draft with ads switched off, it stays unpublished — that is the rule that removed hundreds of URLs from this domain rather than keeping them for a bigger page count.
Layout that does not lie
Primary content first. The H1 matches the query the page was written for. The first eighty words answer it. Ads sit below that answer, not on top of it.
Label ads as ads. Do not style a unit so it reads as a download button, a "next page" control, or a navigation item. If a reader has to guess whether they are clicking your content or your inventory, the placement is deceptive regardless of intent. The same applies to a unit placed directly under a heading so it looks like the section body.
In-article units belong between finished sections, never inside a paragraph or between a heading and its first sentence. One unit per two or three complete sections is enough on a long guide. On a short page, one unit below the answer is the whole plan.
Mobile is where most placements fail. One column, readable type, no interstitial that requires a miss-tap to dismiss, and no sticky anchor that covers the last line of a paragraph. Open the page on a real phone, not a desktop emulator, and read it end to end. If you cannot finish a sentence without moving something, neither can a reviewer.
Reserve space so units do not move the page
Layout shift is a revenue problem before it is a Core Web Vitals problem. A unit that loads late and pushes a paragraph down produces two things: a reader who lost their place, and a click that was probably accidental.
Give every slot a fixed container with a minimum height that matches the smallest ad size you allow in it, and let the unit fill that box. The space is reserved whether or not the unit fills; that is the point. Do not let the ad script size the container after load.
Lazy-load anything below the fold. The first-screen unit, if you run one at all, should be the only one that competes with your text for initial bandwidth. If vitals collapse when units are enabled, the answer is fewer units and a reserved layout, not a different network.
Then measure with units on. A vitals number taken from a clean build with ads disabled is a number about a page your readers never see.
The content bar before you apply
Word count is a floor, not a strategy. What a reviewer actually looks for is whether the page could have been written by someone who has done the thing.
Unique H2s that name a step, not a mood. A dated example with a real currency and a real date. A stated limit — "this is not tax advice", "measured on a small VPS, your numbers will differ". A conclusion that tells the reader what to do next rather than restating the intro.
Locales stay native. Do not fan a 120-word English stub into French, German, Spanish and Russian and call it international coverage. Either a native writer produces a real guide for that market or the URL does not exist. Half-translated stubs are the fastest way to turn one thin page into five.
One URL per topic. Sibling slugs with the same title split their own signals and read as a farm from the outside. There is one AdSense practices URL on this site; everything that was a near-duplicate is a draft.
Placement per template, not per site
Different templates deserve different inventory, and treating them the same is a common self-inflicted wound.
On an article, the answer comes first, then one in-article unit below a finished section, then optionally one more further down on a long guide. On a category or hub page, the content is the list itself — a unit between rows breaks the scan pattern and earns very little. On the homepage, the job is routing people into articles; units there compete with your own navigation.
Utility pages — search results, empty states, error pages, checkout, thank-you pages — get no units at all. There is no query intent to serve, and an ad on a 404 is the clearest possible signal that the page exists to host inventory.
The reviewer screenshot test
Open the URL in a clean profile: no extensions, no logged-in session, cookie banner in its default state, on a phone-sized viewport.
Screenshot the first screen. If what you see is chrome, a cookie wall and an ad, you fail even if the CMS has a thousand words below the fold. A cookie wall that blocks the article before a unit even loads fails the same way a covered H1 does. If the page is readable without measurement cookies, do not trap the first view.
Then scroll once and screenshot again. If the second screen is still not answering the query, the article is padded and the units are carrying it.
Read Search Console before you touch the ad code
Pull the query report for the URL. If a page you wrote about checkout tax display is ranking for "best vpn 2026", the title lied and the traffic will never convert against your inventory. Retitle to the intent you actually served, or noindex the page and rewrite it. Adding units to a mismatched query buys a review, not revenue.
Compare pages by revenue per session, not by impressions. A hub of forty thin pages that each earn nothing will always look busier in the impressions column than four guides that answer a real question. Deciding by page count is how the farms get built.
When a page's clicks come almost entirely from one referral burst rather than search, treat that as a warning about the source, not a win.
AdSense and affiliates on the same URL
They can coexist if both are honest. One disclosure sitting next to the offer, not buried in a footer. Ads labeled as ads. The article intact when you strip both. If a page needs the affiliate block to be worth reading, it is an ad, and it should be marked as one.
What breaks is the stack: an auto-ad above the fold, a "Recommended" affiliate box under every heading, and a newsletter modal on exit. Any one of those is a choice; all three together is a network wearing a blog's template. One call to action per page.
The one-URL audit
Do this on a single live page today rather than planning a site-wide sweep.
Disable your units. Read the page from the top on a phone. If nothing useful remains, set it to draft and stop — no placement fixes an empty page. If a real procedure remains, fix the H1 so it matches the query and rewrite the first eighty words so they answer it.
Now enable one in-article unit below the first finished section, with a reserved container. Reload and screenshot mobile. If the unit covers a paragraph or moves text after load, fix the container before adding anything else.
Check the query report a week later. If the queries match what you wrote, add at most one more unit further down. If they do not, retitle rather than monetize harder.
Repeat on the next URL. Ten pages fixed this way are worth more than a hundred pages of inventory hanging off stubs.
FAQ
Can I run AdSense on a 200-word stub? Not on this site. Stubs stay drafts. A floor of eight hundred words is the automated gate here; the editorial gate is whether the page teaches a procedure.
Does a privacy policy cover this? No. A policy is a legal requirement about data. Everything above is about layout and inventory quality, and the two are reviewed separately.
Auto ads only — is that safer? You still own the screenshot. If auto ads cover your H1 on mobile, turn the density down or give the template explicit slots. "The script decided" is not a defense.
Should I remove ads from old thin posts or delete the posts? Unpublish the posts. A thin page with the ads removed is still a thin page competing with your good ones.
What about VPN and hosting roundups? They stay unpublished here. Wrong intent for this audience and a duplicate of a thousand better-linked pages.
Nine years in software — backend and frontend with equal weight, not a slogan. I also spent a year in IT operations, the kind of work that makes you respect uptime.
The same hands that ship APIs come from graphic design through modeling, 2D and 3D. I care how a thing looks and how it holds together.
Favokres is my personal digital hub: AI-assisted publishing, SEO, a forum, and a shop. An autonomous brain keeps the site moving so pages stay useful, not frozen.
