More likely than not, they are talking about selling your browsing history to major ad networks that already have ads embedded on the sites you use rather than injecting their own ads into others' content. I assume this to be the primary difference between the "snooping" and "non-snooping" plans, because many ISP's already internally keep some amount of browsing history but don't sell it to third parties.

If they are doing ad injection, it is unclear whether or not they could be successfully sued. Some browser plugins with very large user bases inject/modify ads, and although some have received cease and desist letters from content providers, and some have been sued and settled out of court before trial, I don't believe any have gone to trial and lost. The law is murky at best on that issue, and as the ISP, odds are that AT&T would be in a far stronger legal position than even the browser plugins that have been doing this for years.

Not only can they sell your browsing history, they are in the unique position to associate your current IP address with your browsing history, and sell real-time access to that information. If I open up an Incognito window right now and go to a site I've never visited before, they couldn't serve me targeted ads, because without cookies they'd have no reliable way of knowing that I was the same 111.222.111.222 that connected a few days ago. But now, AT&T could set up a server that ad networks can call out to, that can return a browsing profile if queried with an IP address. You've given them permission to do that. As far as I know, Tor is the only way to circumvent this... and with the forthcoming FCC rules, Tor traffic could be significantly rate-limited. This is a pretty big deal.

This highlights the importance of obfsproxy development for tor. Not only can it allow tor's use in regimes that employ deep packet inspection to prevent tor traffic, it could also prevent throttling of tor traffic, as it would look just like normal traffic.

A plain old VPN could also circumvent this.

I actually think that AT&T would be in a weaker position than the browser plugins.

AT&T is a message relayer, while the browser plugin is a specific program installed by the user to modify the messages they receive.

There's no legal question I could hire a secretary to open my mail for me, and white-out the lines I told her I didn't want to see, but it would be a little strange for a mail carrier to open messages and insert ads.

If a user pirates content on the internet, the ISP is not seen as complicit because it is just a dumb pipe, am I right?

Would that still hold if they inserted ads?

Without consent, you're absolutely right. But if (as in the case) the ISP not only obtains consent but also gives significant discounts in exchange for it (and owns the pipe through which the content is delivered, which may give them rights to modify data that a browser plugin company doesn't have), my guess is that AT&T would be in a better position.

The only redeeming feature is that AT&T is giving a discount for the service. I suspect they'd run in to more lawsuits about their monopoly status if they forced that decision on all consumers.

However, the plugin companies aren't in an ambiguous situation at all: someone who requests the page is explicitly requesting it to use the content to run a program which renders an image from it, and is under no contractual obligation (from merely making an HTTP request) to view the entirety of the content. They can neither complain that it's being processed (the browser inherently does that, to make an image from the text), and they can't complain that the end-user is failing to uphold their end of an agreement.

Because it's a user chosen process, and because of how content on the internet is relayed, there's no reason that the plugins are on any kind of shaky ground. (One notable exception would be interacting with dynamic content to bypass a paywall. There likely are other exceptions, but these come down to bypassing some kind of mechanism, not merely failing to load or render a particular portion of the sent content.)

In the article there's a screenshot that strongly implies the ads will be coming via email ("this consent supersedes any prior email unsubscribe notification I may have with AT&T")

I have never received an email from att, though they may getting caught in gmail's spam filter.

I have also been unable to discern ATT's gigapower ads from other ads.

So are you saying that you lack the ability to discern AT&T's GigaPower ads from other ads, or are you saying that AT&T's ads are indiscernible from other ads?