Types of Advertising and Digital Ad Types.

There is no ideal ad setup. The question that actually matters is which combination of formats prices this audience without shrinking the session.

Independent web businesses — content sites, tools, games, apps — still get asked the same things: where the units should go, how many display versus native, whether interstitials are too much, whether rewarded is an ad or a product decision. Those are the wrong first questions. Formats dilute one another. Adding one more unit can raise page RPM and cut visits, time on site, and return rate at the same time.

This piece walks the formats that still matter, the ones that mostly don’t, and the test you should run instead of hunting for a permanent layout.

There Is No Permanent Answer.

Devices, browsers, demand, and attention change. A layout that worked last quarter is a hypothesis, not a setting.

Add to that ad blindness, seasonality, and shifting traffic sources and the target moves again. The useful fact underneath every format debate:

All ads dilute one another.

A second unit does not add its own revenue on top of the first. It competes for the same attention. The right test is what happens to the whole visit — and to the next one — not what happens to a single placement.

Measure the Session, Not the Page.

Page RPM can rise while income falls. One more in-content unit might win the auction on that page and still cost you the second pageview, the return visit, or the user who would have opted into a rewarded exchange later.

Use earnings per session — what a visit is worth — as the success metric. If you want the longer argument for why page RPM is the wrong north star, read why RPM is not the best metric for optimizing ad revenue.

Display Ads.

Display is still the backbone — responsive units in the page, priced in a real-time auction. AdX is the exchange most stacks still clear through. Amazon, Magnite, PubMatic, OpenX, and Index Exchange are the other names you will see in a serious header-bidding stack. Ad networks still exist; most of them buy those same exchanges.

Direct deals and PMPs belong in that auction so reserved inventory still has to beat the open bid. A direct deal helps when it wins. It does not help when it merely fills the slot and takes it out of competition.

Affiliate units are a different revenue model (cost per action), not a display creative. They can sit next to ads if you place them so they do not fight the same click. They are the wrong default for a media site that is not writing for an advertiser’s catalog.

Native Ads.

Native units are built to look like the surrounding content — a card in a river, a recommended-story row. They can work, especially on social and discovery traffic. They also send people away from your site, usually on a CPC.

Test native against display on the same session metric — do not assume one wins. A combination can be worth more than either alone. If session value is already high, sending the reader to someone else’s article is an expensive click.

In-Content Placements.

Older “in-line” ads meant highlighted words that popped a unit on hover. That format is mostly gone — blockers, mobile (no hover), and policy all worked against it.

What remains is in-content display: a unit in the article flow, between paragraphs. Same auction as sidebar display, different attention cost. Use it in moderation and judge it on the session, not the slot.

Anchor Ads.

An anchor (or overlay) sits at the bottom of the viewport and stays there. High viewability, which is why demand pays for it. It is also always on screen, which is why it can tax a session if density is already high.

Interstitials.

A full-screen unit between page views, levels, or screens, with an obvious close. On content sites they are easy to overuse: bounce and time-on-site move first. On web games they belong at natural breaks — between levels, not mid-play — where players already expect a pause.

Video — Sticky and In-Stream.

Sticky video follows the reader and stays viewable. In-stream ad breaks (pre-roll, mid-roll) sit inside a video the visitor already chose to watch. Video is often the dearest demand on a page that can support it. Measure the session — these units still dilute everything else on that page.

Rewarded Ads.

Rewarded is not another unit. The visitor opts in: watch a short ad, unlock the next thing — an article, an export, a continue, another life. That is a revenue-model choice, the one mobile apps have used for years and most of the web still treats as a format experiment.

It belongs anywhere the product has something worth gating. Games and tools feel it first; content sites use it to keep reading or skip a wall. Judge it on whether the exchange is honest, not on whether you can staple it next to five display units.

App Formats.

In-app inventory is not AdMob dropped into an old ad server and forgotten. Anchor, native, and interstitial units have to sit at product moments — between screens, not over the task. The auction is the same idea as the web; the placement rules are not.

Formats That Usually Fail Now.

Pop-unders are largely a dead format: in-browser blockers, weak demand, and users who treat them as hostile. In-text and rollover keyword ads belong in the same bucket. If a format needs to hide to get the click, it is not a strategy.

Ad Types and Density Have to Be Aligned.

Before you add a format, decide what you are willing to trade. Some tests spend a little experience for a lot of session value. Some protect experience and still raise the session because the mix got less noisy. “Add one more” is not a strategy; it is how pages get loud and visits get short.

The search space — format × position × density × device × traffic source — is larger than a person can run by hand. That is what an optimization layer is for. The human job is the constraint: what you will not sacrifice.

What to Do Next.

Learn the formats. Put them on the page as hypotheses. Measure the session. Keep the ones that raise what a visit is worth without training people to leave.

First published December 2015. Rewritten August 2026.