01
One h1, phrased as the buyer's question
Every page answers exactly one question, and the h1 is that question in the buyer's words. Machines treat the h1 as the strongest signal of what a page is for. One page, one question, one h1.
Why scrape bias
Extraction has a bias. Build for it.
Answer engines do not read a page the way a person does. They crawl it, extract candidate passages, and assemble an answer from the passages that survive. That process favours pages that answer first, structure cleanly and mean exactly one thing. A page that buries its point under three screens of brand prose may still rank in a list of links; it rarely survives extraction intact.
Ambiguity is a tax. Every place a page makes the machine guess, two names for the same service, a claim that only renders after JavaScript runs, an FAQ hidden inside an accordion the crawler never opens, is a place the answer can come out wrong, or come out citing someone else. The standard below removes the guessing.
The standard
Eight rules, adoptable today.
Nothing here needs new tooling or a replatform. Each rule stands alone and can be applied to one page at a time, with the team you already have. Together they make a page that reads well for a person and extracts safely for a machine.
02
An answer-first opening
The first paragraph under the h1 answers the question fully, on its own, before any context or positioning. Assume the extractor takes that paragraph and nothing else. If it survives alone, the page survives extraction.
03
Semantic headings and real text
Structure lives in h2 and h3 elements, not styled divs, and the content is real text in the markup. Nothing load-bearing sits behind JavaScript: if a claim only exists after a script runs, assume the machine never reads it.
04
A visible, non-collapsed FAQ
Five questions per page, each an h3 with its answer in plain text below it. No accordions, no tabs, no click-to-reveal. The questions mirror what buyers actually ask, because those are the prompts the engines receive.
05
Valid JSON-LD on stable URLs
Service or product schema plus FAQPage, validated, on production URLs with shallow paths that do not change. Structured data hands the machine facts instead of prose to interpret; URL churn throws those facts away.
06
One clear CTA
Each page ends with a single next step. Competing calls to action read as noise to a person and as ambiguity to a machine summarising what the page wants the reader to do.
07
An llms.txt index at the domain root
A plain-text file that states what the organisation is in one line and lists the canonical pages, so a crawler on a token budget reads the curated version first. Ours is live at /llms.txt.
08
Entity consistency
The same firm name, the same service names, the same one-line descriptions, everywhere: on the page, in the schema, in llms.txt, in third-party profiles. Machines resolve entities by repetition; every variant spelling splits your identity.
Where it connects
Part of one estate.
The standard is the page-level layer of a wider discipline. Here is where it joins the rest of the work.
The wider picture
AEO for Enterprise Estates Field guide
Measure first
DXP Value Survey The front door
AI in the estate
AI and Gen AI Solutions
How answer engines read, what moves visibility across a whole estate, and how to measure it.
A fixed-price four-week diagnostic. Machine visibility is scored inside its AI-readiness dimension.
Agentic migration plus the AI capability built into the live experience your customers touch.
Questions
Frequently asked questions.
What is scrape bias?
The tendency of AI answer engines to favour pages that answer first, structure cleanly and resolve to one unambiguous meaning. Extraction keeps the passages it can lift safely and discards the rest, so the bias rewards pages built for it.
Do we have to adopt all eight rules at once?
No. Each rule stands alone and can be applied page by page. Most teams start with the question-shaped h1 and the answer-first opening on their highest-value pages, then add the FAQ and the schema, then publish llms.txt once the canonical pages are settled.
Does this replace SEO?
No. The standard is compatible with conventional SEO and usually improves it, because search engines also reward clean structure and structured data. It adds the disciplines that ranked search never demanded: answer-first writing and entity consistency.
Why publish an internal standard?
Because standards work better in the open. We would rather the pages machines quote, ours and everyone else's, be built to answer correctly. Publication also keeps us honest: every page on tbscg.com can be checked against this one.
How do we know if a page meets the standard?
Read the first paragraph in isolation and ask whether it answers the h1 on its own. Validate the JSON-LD. View the page with JavaScript disabled and check that nothing load-bearing disappears. Then ask an answer engine the h1's question and see what comes back.
Start here
Apply it to your estate.
Adopt the standard as written; it is published to be taken. If you want to know how far your estate currently sits from it, the DXP Value Survey scores machine visibility as part of its AI-readiness dimension, in four weeks at a fixed price.