AMP: What Google's Speed Play Offers, and What It Asks

Google introduced a new acronym to the industry on October 7, and this one arrives with launch partners, a GitHub repository, and a diagnosis nobody can dispute. The mobile web is too slow.

Accelerated Mobile Pages (AMP) is an open-source format for building web pages that load, in the demos and early tests, close to instantly. A deliberately restricted subset of HTML. A curated JavaScript library replacing arbitrary scripts. And, the piece doing the heaviest lifting, Google’s own servers caching and serving the pages from its edge. When a platform offers you speed in exchange for control, price the control.

Dozens of major publishers are aboard at launch. The positioning against Facebook’s Instant Articles and Apple News is transparent. The platforms are competing to solve mobile publishing’s speed crisis inside their own walls. And the announcement’s subtext for search visibility is not subtle.

We have covered enough platform bargains (the app stores, the social feeds, the Reader era) to recognize the genre. So the useful job here is to separate what AMP genuinely improves from what it asks publishers to give up. State the offer honestly. State the ask honestly. Sort the response by audience, because the right answers here genuinely differ by who you are, and the wrong answer is adopting or dismissing on reflex.

The offer, stated fairly

The speed is real and the mechanism is legible. AMP’s restrictions remove, by fiat, essentially every category of slowness we have spent a decade auditing. No author JavaScript outside the sanctioned library. Static layout declarations that prevent the reflow jank of ordinary pages. Resource loading the framework controls.

Add pre-rendering from Google’s cache when a result is likely to be tapped, and the experience nears instant in a way ordinary hosting cannot match. The page is already there.

The strategic sweetener, plainly implied if not yet formalized: placement. Google is building carousels and treatments for AMP content in mobile search. Publishers signing on early are betting the format’s pages will enjoy visibility ordinary pages do not.

For news operations living and dying on mobile search and social referral, that combination, instant loading plus privileged placement, is a serious offer. The launch partner list says the industry’s biggest players priced it as one.

The ask, stated equally fairly

The bargain’s other column, which the launch coverage under-weights.

A second rendering target. AMP pages exist alongside your canonical pages. That means parallel templates, parity upkeep, and the two-versions-of-the-truth workflow our m-dot coverage spent years pricing. The format’s limits also mean features your ordinary pages have (forms, interactive elements, your analytics stack) exist in reduced or adapted form.

Measurement inside the fence. Analytics in AMP works through the framework’s components, with real limits against the instrumentation our measurement practice installs. Attribution gets murkier when pages serve from a cache you do not control.

Your pages, Google’s URLs. Cache-served AMP pages carry google.com addresses in the browser, which touches brand presence, sharing behavior, and the URL equity our SEO practice husbands. The ecosystem acknowledges the concern, and it is unresolved at launch.

Sum the ask and it is the platform-bargain shape we have catalogued since the Facebook era. Reach and speed, in exchange for format control, infrastructure dependence, and terms set by the party granting the placement. That shape is not disqualifying. It is the thing to price with eyes open, on a bargain whose terms the granting party can revise.

The precedents worth consulting

Because platform bargains reward pattern memory, here are the relevant precedents, read against this one.

The app-store precedent. Distribution granted on the platform’s terms, terms revised repeatedly, and the publishers who built directly on the placement rather than on their own properties absorbed each revision. AMP’s placement sweetener should be enjoyed, if piloted, with that history in the deck.

The Facebook Instant Articles precedent, running in parallel right now. Same diagnosis, same publisher anxiety, same trade of format control for platform performance. The instructive difference: AMP is at least built from open web technology at public repositories. Cache-served but crawlable. That makes it the more hedgeable of the two bets, and the one whose lessons transfer to your own pages.

And the Reader-era precedent, which governs the exit thinking. Platform features that publishers organize workflows around can be discontinued by memo. Any AMP pilot should be built as a derivative pipeline from canonical pages. Templates generated from the system of record, per rendition approach, rather than a parallel authoring habit. That keeps the withdrawal cost at ‘turn off the template’ rather than ‘untangle the workflow.’

None of the precedents forbids the pilot. All of them shape its architecture, which is what precedent files are for. Bargains age. The build-on-what-you-own advice has not yet.

The response, by audience, and the lesson to steal

For news and content publishers whose economics ride mobile search and whose competitors are launching partners: pilot it, seriously and soon. The placement stakes are real and the format was built for your content shape. Instrument the pilot per measurement practice and price the workflow tax honestly.

For the institutional majority of this readership: watch, and decline the reflex. Your program pages, admissions content, and service information are not carousel candidates. The workflow tax buys you little, and the dependency is real.

But do not file the announcement and move on, because AMP’s deepest content is a lesson argued here for years, now demonstrated at Google scale. Nothing in AMP’s speed is unavailable to a disciplined ordinary page.

The format’s restrictions are a performance budget with enforcement. Constrained scripts. Declared layouts. Controlled resources. Every one of those constraints is adoptable voluntarily, on your own pages, under your own URLs, with your own analytics. The sites that adopted the budgets, the third-party diets, and the weight rules of our usual program already load in the territory AMP is selling.

Google just built a walled garden to prove the open web’s own best practices work. Steal the proof. Keep the deed. Let the publishers’ pilots report back on the garden’s terms. We stay open, and the carousel era it predicts will be worth watching from a fast site of your own.