The Complete Guide to Google AdSense Optimization in 2026

Why Accounts That Used To Perform Well Are Quietly Losing RPM

Most of the accounts I audit were configured correctly at some point — usually two or three years ago — and then never touched again while everything around the ad units changed. The template got a redesign. Paragraph length shortened because someone read a guide about "scannable content." A caching plugin update changed how the page loads scripts. None of these are AdSense mistakes on their own, but together they quietly break placements that used to sit in exactly the right spot relative to the fold, the first scroll, or the natural reading pause after an intro paragraph.

The tell is almost always the same: traffic holds flat or grows, but RPM slides a few cents a month for six months straight, and nobody catches it because a few cents doesn't trigger any alarm. By month six, you've lost 15-20% of your per-thousand-impression revenue and the only reason you'd notice is if you happened to pull up a year-over-year comparison. I check year-over-year RPM by device and by page template before I look at anything else, because that's where drift shows up first.

I also check whether the account is running the same block list and category exclusions it had two years ago. Verticals shift — a site that used to publish general lifestyle content and pivoted toward personal finance needs a completely different sensitive category exclusion list, and an outdated one either blocks advertisers who'd now bid well or lets through categories that clash with new content and depress CPMs across the board.

• RPM falling while sessions and pageviews stay flat or grow
• A rising share of ad requests marked unfilled even though overall demand hasn't changed
• Viewability percentage dropping on units that used to sit above 70%
• Ad density that no longer matches current paragraph length or page structure

Choosing The Right Ad Unit Type Instead Of Defaulting To Display

Display ad units are the default for a reason — they work almost everywhere — but treating them as the only option is where a lot of accounts leave money on the table. In-feed ads are built for exactly one situation: a list or grid of content, like a homepage river of posts or a category archive. Dropped into that context, an in-feed unit blends into the layout the way a native ad should, and I've seen click-through rates run 2-3x higher than a display banner wedged into the same sidebar.

In-article ad units are the ones most publishers underuse relative to how well they perform. They're designed to sit inside a block of text and adapt their size to the surrounding content, which matters a lot on long-form pages — a 1,800-word recipe post or a 2,500-word explainer has room for two, sometimes three in-article units without feeling stuffed, as long as you space them by scroll depth rather than by a fixed pixel count.

Matched content is the one people either ignore entirely or overuse to the point of hurting the reader experience. Used well, on a site with 40+ published posts, it recirculates readers into more pageviews per session, which raises RPM indirectly even before you count its own ad slots. Used badly — placed above the fold, or on a site with only a handful of posts so it recommends the same five articles everywhere — it just adds clutter without lifting session depth at all.

Responsive display units are the safer default over fixed sizes now, simply because they adapt to whatever container width the template actually renders at across breakpoints, and a fixed 336x280 unit crammed into a container that resizes on tablet width just gets awkward padding around it. The one place I still hardcode a fixed size is the above-the-fold desktop placement, because a 728x90 leaderboard at a known width outperforms a responsive unit that might render smaller on that exact slot.

The Placement Hierarchy That Actually Moves RPM, Not Just Fill

The three placements I mentioned in every audit — in-content after the second paragraph, a mobile anchor unit, and a unit sitting just above the fold — are still the highest performers I see across verticals, and that hasn't changed. What's changed is the fourth and fifth placements worth adding once those three are dialed in: a second in-article unit after roughly 60% scroll depth on anything over 1,200 words, and a sticky sidebar unit on desktop that only fires once the sidebar has fully entered view.

Reserve the layout space before the ad loads, every time, with a fixed-height container sized to the expected unit — this is the single fix that prevents the most common self-inflicted problem I see, which is layout shift tanking both Core Web Vitals and the auction's willingness to bid high on that impression. If you haven't audited this site-wide, it's worth running through a page speed and Core Web Vitals checklist alongside your ad placement review, because the two problems usually travel together.

Density matters more than any individual placement decision. A page with six ad units and 900 words of content isn't going to out-earn a page with four units and 1,600 words, because the algorithm reads the extra units as low-value inventory and prices them accordingly. I'd rather see three well-placed units on a shorter post than five units chasing every available slot Auto ads suggests.

• In-content unit after the second paragraph
• Mobile anchor unit
• Unit above the fold on article pages
• Second in-article unit at roughly 60% scroll depth on long-form content
• Sticky sidebar unit on desktop, fired only when in view

What Auto Ads Is Actually Optimizing For, And Where It Breaks

Auto ads isn't randomly sprinkling units across your page — it's running a prediction model against your historical viewability, click-through, and session data to estimate which slots will produce the highest revenue per pageview, then placing units at the density setting you've chosen. That's genuinely useful as a baseline, especially on a new site with no placement history to work from, because it will usually find two or three spots you wouldn't have picked manually.

Where it falls apart is context it can't see. Auto ads doesn't know that the table in your comparison post is the actual conversion point of the page, so it'll happily drop a unit inside it. It doesn't know your cornerstone content is meant to build authority rather than maximize ad density, and it doesn't distinguish a 400-word news brief from a 3,000-word guide when deciding how many units to insert — it just runs the same density logic against both.

My default recommendation is manual placements for anything on a cornerstone or comparison page, Auto ads set to medium density everywhere else, and a regular pass through the ad review center to block individual auto-placed units that are underperforming or sitting somewhere they shouldn't. Once a site is consistently clearing $10,000-15,000 a month through AdSense, it's also worth running the numbers on how Ad Manager compares to AdSense for header bidding — Auto ads has a real ceiling, and that's usually where you hit it.

Testing Placements Methodically Instead Of Guessing

AdSense's built-in experiments feature is limited — it'll test a handful of dimensions like ad balance or a couple of placement variants, but it won't let you run a true controlled test across arbitrary layout changes. The workaround I use is a cohort test: pick two sets of pages that are as similar as possible in topic, length, and traffic source, change one variable on one set, leave the other alone, and compare RPM over a full two-to-four week window rather than a few days.

Two to four weeks matters because day-of-week and traffic-source mix swing RPM by 10-15% on their own — a Tuesday-to-Wednesday comparison will lie to you constantly. I've had clients kill a placement change after three days because RPM dipped, only to find out it was a weekend organic traffic dip unrelated to the ad unit at all. Don't call a test until you've got at least 20,000-30,000 impressions per variant, and don't test more than one variable at a time or you won't know what actually moved the number.

• Pick comparable page cohorts before changing anything, not before-and-after on the same pages
• Run for 2-4 weeks minimum, not days
• Isolate one variable per test: one new placement, one density change, one unit type swap
• Require 20,000+ impressions per variant before drawing a conclusion
• Track page RPM and viewability together, not RPM alone

Mobile And Desktop Aren't The Same Optimization Problem

On most publisher sites I look at, mobile is 65-80% of sessions, so mobile ad performance basically is your ad performance whether you've treated it that way or not. The anchor unit is the highest-leverage mobile placement because it's persistently viewable without depending on scroll behavior, but it's also the one most likely to get penalized by Google's own page experience signals if it covers more than 20% of viewport height or can't be dismissed. I check anchor unit coverage on a real device, not a browser dev tools emulator, because rendering differences are common enough to matter.

Desktop behaves differently: sidebar units get real viewability because desktop readers scroll less relative to content length, and larger ad sizes — 300x600 instead of 300x250 — consistently outperform on desktop specifically because there's real estate to support them without crowding the content column. A unit that performs well in a desktop sidebar will often underperform badly if you force the same size into a mobile layout, so resist the urge to treat "responsive" as "identical" — the two device types need separate placement logic, not just a flexible container.

• Anchor unit under 20% of viewport height and dismissible
• In-article units spaced by scroll depth, not fixed word count, since mobile line-wrap changes effective word density
• Sidebar units disabled entirely on mobile rather than stacked awkwardly under content
• Larger ad sizes (300x600) reserved for desktop breakpoints only

The Seasonal RPM Swings That Catch New Publishers Off Guard

RPM isn't flat across the year, and if you're comparing November to February and concluding your account is broken, you're comparing the wrong months. Q4 — specifically the back half of October through December — carries the highest advertiser demand of the year across almost every vertical, driven by holiday retail budgets, and it's normal to see RPM climb 30-50% above your summer baseline during that window. January and February are the mirror image: budgets reset, demand drops, and a 20-25% RPM dip against December is completely typical, not a sign something's wrong.

Vertical matters here too. Finance and personal finance content sees a secondary bump around tax season in Q1 in the US market. Travel content softens in the deep winter months outside of holiday booking windows and picks back up hard in spring. Education and back-to-school content spikes in August and September. If your content calendar can shift even 10-15% of publishing volume to land more pieces right before your vertical's peak season, you're getting paid more for the same amount of writing.

• Q4 (Oct-Dec): highest RPM of the year across most verticals, often 30-50% above summer baseline
• January-February: expect a 20-25% dip as ad budgets reset — don't panic-test placements during this window
• Finance content: secondary bump around Q1 tax season
• Travel content: soft in deep winter, recovers in spring
• Education content: spikes August-September

Reading The Reports UI As A Diagnostic Tool, Not A Scoreboard

Most publishers open the AdSense reports screen, look at the total revenue number, and close the tab. That number tells you almost nothing about what to fix. The useful workflow starts with page RPM broken out by URL or template, compared against impression RPM and match rate for the same segment — if match rate is high (95%+) but RPM is low, your problem is demand quality or placement, not fill. If match rate itself is low, you're looking at a technical or policy issue blocking requests before they ever reach the auction.

Active view viewability is the metric people check least and should check most. A unit sitting at 45% viewability is getting bid on as if it's rarely seen, because it usually isn't — and no amount of placement optimization elsewhere will fix a fundamentally invisible ad slot. If you're unclear on how these metrics relate to each other, it's worth a slower read through how RPM actually gets calculated and what moves it before you start changing placements based on a metric you're misreading.

The other habit worth building is segmenting by device and by referral source before you conclude anything site-wide. I've seen accounts where desktop RPM looked healthy at $12 while mobile sat at $4.50, and the blended number in the top-line report masked a mobile-specific problem entirely — in that case it was a mobile anchor unit that had silently stopped serving after a theme update three months earlier.

• Match rate 95%+ with low RPM = demand/placement problem, not fill
• Match rate below 90% = technical or policy issue blocking requests
• Viewability under 50% = placement is effectively invisible, fix before anything else
• Always segment by device and referral source before concluding site-wide

Content Depth, Policy Risk, And Knowing When You've Outgrown AdSense

There's a real, measurable relationship between content depth and RPM that has nothing to do with keyword targeting. A 600-word post structurally supports one, maybe two ad units before density looks aggressive. A 2,000-word guide supports four or five, naturally, because there's enough reading time and scroll distance between them. Thin content isn't just an SEO problem — it's a ceiling on how much ad inventory a page can support at all, regardless of how well you optimize placement within it.

AI-assisted content published without real editorial review is the policy risk I see trip up more accounts than anything else right now. It's not that AI-assisted writing is banned — it isn't — it's that unreviewed output tends to be shallow, occasionally wrong, and thin in exactly the way that both hurts RPM per the point above and increases the odds of a policy flag on originality or usefulness grounds. Have an editor actually read it before it publishes, not just run it through a checker.

Consent management setups deserve a direct check, not an assumption that your CMP vendor handled it correctly. In restricted regions, ad requests need to actually wait for consent before firing — plenty of setups technically show a consent banner while ad tags fire underneath it anyway, which is a compliance problem, not just a revenue one. Re-check your ads.txt file after any change to monetization partners too; a stale ads.txt silently suppresses demand from partners who can no longer verify you as an authorized seller.

I've watched this play out directly: a client rewrote a batch of 500-700 word posts into 1,500-1,800 word versions with added FAQ sections and comparison tables, without touching the ad setup at all, and per-page RPM rose 22% purely because the extra length supported an additional in-article unit and pushed the matched content block further down into higher-intent scroll territory. No new placements, no new traffic — just more page for the existing ad logic to work with. Once you're consistently earning $8,000-10,000+ a month and you've fixed placement, density, and technical issues, the next lift usually doesn't come from AdSense at all — it comes from qualifying for additional demand sources layered on top of it. Before assuming you need to rebuild your stack, it's worth running a quick eligibility check for premium demand programs to see what you already qualify for with your current traffic and content profile.

Frequently Asked Questions