Audit the web
that AI reads.

OpenAI, Anthropic, Perplexity look here for crawl preferences. Without it, AI crawlers default to whatever robots.txt says — which usually says yes.

AI Overviews lean on byline + Schema.org Person markup to attribute and trust a passage. A page with no author entity is harder to cite.

Pages thick with named entities — products, people, places, dates — are easier for retrieval-augmented systems to index against a query.

FAQ + Article structured data are the cleanest way for an LLM to extract a Q&A pair verbatim. Without it, the model has to guess.

Article markup gives the engine date, author, body — the metadata that ends up in the citation card next to your answer.

Pages that already cite their own sources signal credibility. Engines that cite *you* prefer pages that *cite others*.

An island page with zero outbound links to authoritative sources reads, to a model, like an unverifiable claim.

Even with outbound links, lack of in-text citation cues ("per", "according to", "source:") makes machine extraction noisier.

For voice + assistant surfaces, the SpeakableSpecification schema marks the passage that should be read aloud.