AI crawlers now read the web at scale, and their summaries sometimes answer the user without a click — your research, your phrasing, zero visits. Publishers face a genuine strategic fork: block the bots and preserve leverage, or feed them and bet on citations. The right answer depends on what your site sells.
The case for blocking
If your business is pageview-funded — ads, paywalls, subscription content — uncredited consumption is a cost: the summary satisfies the query, the visit never happens, the impression never renders. Blocking the AI crawlers (robots.txt rules targeting the known bots) preserves the scarcity that funds you. The counter-risk: assistants become the interface for informational queries, and invisible sites eventually lose even the search-mediated traffic as user habits migrate.
The case for feeding
If your business is authority — services, products, leads, brand — being the cited source is marketing. A Perplexity answer quoting your research delivers pre-qualified visitors who trust you on arrival, and the citation compounds into the "known entity" status every source-based strategy needs. The counter-risk: the citation economy is young, unmeasured, and controlled by platforms that change their minds.
The middle path most sites should walk
Split by content class. Feed the machines the content that markets you: methodology pages, original data, definitive guides — the citable assets. Block (or leave ambiguous) the content that monetizes per-view: your ad-supported long tail. This is enforceable at the directory level in robots.txt and mirrors what paywalled publishers do with their free/premium split. Pair it with measurement: log AI bot hits, watch assistant referrals, and review the split quarterly — this policy should follow evidence, and the evidence is only starting to exist.
Whichever side you pick, decide deliberately. The default — do nothing — is itself a policy, just not one you chose.
