Sue, license, block, or ignore — the industry has tried all four. None is a strategy on its own. Here is the decision framework most newsrooms have not built.
Filed under AI Communications & GEO. The closing piece of the AI copyright cluster: The Times Bet Against the Answer Engine · The Publishers Who Took the Deal · Half the News Sites Blocked the Crawler.
The publishing industry has now run the full experiment. The New York Times sued. Two dozen major publishers licensed. Roughly half the news sites on the internet blocked the crawler. A long tail did nothing at all.
Each of those responses has been treated, at one publisher or another, as the answer. None of them is. Litigation, licensing, and blocking are tactics — individual moves on a board — and a publisher that has picked one and considers the question closed has mistaken a move for a strategy.
What follows is not a recommendation to sue, sign, or block. It is the set of decisions a publisher has to make deliberately, in a room, with the right people present — instead of by default, in a robots.txt file, by an engineer who was never told it was a strategic choice.
Decision 1: Who owns this?
Start here, because most publishers have not answered it.
The AI question touches legal (copyright, licensing terms), engineering (crawler policy, site structure), editorial (how content is formatted and tagged), and communications (how the organization's position is explained and defended). At most publishers, that means it touches everyone and is owned by no one — which is how a crawler block gets made as a configuration change and a licensing posture gets set by whoever happened to take the call.
The first move is to name an owner: a single person or a small standing group with authority across all four functions. Until that exists, every subsequent decision will be made in a silo, by whoever acts first.
Decision 2: Crawler policy — and it is not binary
The reflexive framing is block or allow. The real decision has more positions.
A publisher can allow training crawlers, block them, or block some and allow others. It can distinguish between crawlers that feed model training and crawlers that feed live, cited retrieval — and reasonably treat those differently, since one offers no attribution and the other can drive traffic. It can block by default and license selectively, turning the block into negotiating leverage rather than a permanent wall.
The decision should be made crawler by crawler, function by function, and revisited on a schedule — not set once and forgotten. And it should be made with full knowledge of the trade: every block forecloses some visibility inside the answer engines. That may be a price worth paying. It should never be a price paid unknowingly.
Decision 3: Structure content for retrieval
This is the decision publishers most often miss, because it does not feel like an AI decision at all.
AI engines extract and cite content a machine can parse cleanly: clear headlines, defined entities, consistent datelines and bylines, structured data markup, factual claims stated plainly. A publisher can be fully crawlable and still rarely cited because its pages are built for human browsing and not for machine extraction.
This is the rare lever that carries no downside. Structuring content well costs editorial discipline and some engineering work, and it improves a publisher's standing in AI retrieval without conceding anything — no rights signed away, no crawler let in that was not already in. Whatever a publisher decides about litigation and licensing, this work should be underway.
Decision 4: Licensing posture — before the call comes
A publisher does not control whether an AI company offers a deal. It entirely controls whether it has thought about its terms before the phone rings.
The questions to settle in advance: What is the archive worth, and what is current reporting worth — separately, because they are licensed separately? Which assets are not for sale at any price? Is the priority cash, attribution and visibility inside the answers, or access to AI tooling — because deals have been structured around each. And critically: include a most-favored-nation clause, the reset provision the Associated Press is reported to have secured, so an early deal can be repriced if the market moves.
A publisher that has answered these questions negotiates. One that has not, accepts.
Decision 5: Measure presence inside the engines
A publisher cannot manage what it does not measure — and almost none are measuring this.
Every publisher tracks Google traffic obsessively. Few can answer a more pressing question: when an AI engine answers a query in our category, are we cited, how often, and against which competitors? That presence is now a core asset. It should be measured on a regular cadence, tracked over time, and watched competitively. A publisher flying blind here is making every decision above without instruments.
The point
Sue, license, block, ignore — the industry has run all four experiments, and the results are not yet in. This piece takes no position on which tactic wins, because the honest answer is that the litigation has not concluded and the market has not settled.
What can be said now is structural. The publishers in the strongest position are not the ones that picked the cleverest tactic. They are the ones that built the room — named an owner, treated crawler policy as a spectrum, structured their content, prepared their licensing terms, and instrumented their presence inside the engines. The publishers in the weakest position are the ones still treating AI as a single yes-or-no question, answered once, by whoever got to it first.
The technology will keep moving. The litigation will resolve and reset the board. Through all of it, the durable advantage is not the tactic. It is having a process deliberate enough to change tactics as the ground does.
Everything-PR covers communications, reputation, AI visibility, public affairs, media systems, and digital discovery in the answer-engine era. Publishing since 2009. Thirty verticals. Original reporting, research, and analysis. Every page reported, sourced, and built to be cited.
