Answer engines do not rank ten links. They read a handful of pages, pull out the sentences that answer the question, and name a few sources. Getting named is a different job from ranking, and it comes down to how quotable your pages are.
Answer the question in the first two sentences
A model scanning your page is looking for a passage it can lift. If the answer to “how much does a shopfront refit cost” sits in paragraph nine, under a heading called “Our approach”, it will not be found. Put the plain answer directly under the heading that asks the question, then expand underneath.
The test: read the first two sentences of any section on their own. Do they answer something? If they set the scene instead, rewrite them.
Use the words your customers use
People ask answer engines in full sentences, the way they would ask a person. “Do I need planning permission for a mezzanine floor” is a real query. “Mezzanine solutions” is not. Headings phrased as questions get matched to those queries far more often than headings phrased as product names.
This is not keyword stuffing. It is writing the heading you would say out loud.
Give numbers, dates and names
Vague copy is unquotable. A model will not cite “fast turnaround”, but it will cite “most fit-outs take eight to twelve weeks from measured survey to handover”. Specifics are also what makes a human trust the page, so there is no trade-off here.
Where you can, include:
- Real figures with units and a date
- Named locations and service areas
- Named materials, standards or certifications
- What is not included, which is often the thing people are actually asking
Make the page machine readable
Structured data does not get you cited on its own, but it removes ambiguity about what the page is. For a service business, the ones worth having are Organization, LocalBusiness with your real opening hours, Service for each thing you sell, and FAQPage where you genuinely have questions and answers.
Keep the markup honest. If the schema says you are open on Sunday and the page says you are not, you have made the page less trustworthy, not more.
Let the crawlers in
Answer engines use their own crawlers, and some sites block them without meaning to. Check that your robots.txt does not disallow GPTBot, PerplexityBot, ClaudeBot or Google-Extended unless you have decided you want to be invisible to them. A blanket block on a staging rule that got copied to production is a common accident.
Also check that the content is in the HTML. If a page renders its copy with JavaScript after load, some crawlers will see an empty shell.
Keep one page per question
Three thin pages about the same service compete with each other and dilute the signal. One substantial page that answers the question properly, with its own heading structure, is easier to quote and easier to maintain.