Could AI answer your customers’ questions?

What actually happens when you run it

This isn’t a checklist or a crawler score. We read your site with the same scraper our paying customers use, build a real Holp assistant from what we find, and then interview it: the questions a first-time customer asks before buying, booking or visiting — phrased exactly the way a person types them. “How many guests can we have?”, not a string of keywords. Every answer comes back with the page it came from and the line on that page that backs it up, so you can see the assistant’s working. Then the assistant stays on the results page and you can ask it anything you like — five questions free, no account needed.

Because it is the real product rather than a demo of one, the result is honest in both directions. Where your content is good, you will watch an assistant answer a stranger’s question accurately from your own pages. Where it isn’t, you get a specific, fixable list.

Three outcomes, and what each one costs

Not on your website. A customer asking this gets nothing — so they ring, email, or go elsewhere. That’s a lost booking, sale or enquiry you never hear about: you know about the calls you get, not the people who left without asking. With Holp on your site they don’t disappear — it takes their details and tells you exactly what they asked.

On your website, but not written as an answer. The information is there, but it’s written as description rather than as a direct answer, or it sits several levels into your site — so a customer skimming, or an AI reading your pages, has to work it out. Holp answers these directly from your existing content — with the event card to book, the product to add to basket, or the link to enquire, right there in the conversation — without you writing anything new.

Answered clearly. Your website states it plainly and a customer would find it. Nothing needed.

Why the gaps are the valuable part

We found this out by accident. Customers setting Holp up would test it the way anyone tests a new tool — firing their own questions at it — and kept coming back with a version of the same sentence: “it couldn’t answer that, and actually neither can our website.” The assistant hadn’t failed. It had just been the first thing to ask their site a real customer question and report back honestly.

That turned out to be worth as much as the automation. Teams were rewriting FAQ pages, adding parking and access details, and publishing the pricing questions their sales inbox answered fifty times a week. Every one of those answers is a customer who no longer has to ring to find out — and every one also helps the humans reading the site, the search engines indexing it, and the AI assistants now answering questions about you whether you are involved or not.

This tool is that discovery, made deliberate. You do not have to sign up and wait for the pattern to emerge — you get the list in one pass, and our knowledge gap tool keeps finding new ones from what your customers actually ask once Holp is on your site.

Reading your results honestly

We read a sample of your site, not all of it, starting with the pages customers reach first, and the report tells you exactly how many pages that was out of how many you publish. So a question marked as not on your website means “not on the pages we read” rather than “nowhere on your site” — which is why the assistant can also search your live site when you chat with it.

One thing we deliberately don’t test: whether you show up in AI search results, or whether ChatGPT recommends you. This check is about your own website — whether the answers your customers need are on it, written in a form a person or an AI can actually use.

Treat the gaps as a content backlog rather than a verdict. Most are a paragraph on a page you already have. Answer them and everything downstream gets better, whether or not you ever use Holp.

What we do with your site

We respect robots.txt, identify ourselves honestly as HolpAnswerabilityBot, and read a bounded number of pages. The assistant we build for you is private and is deleted automatically if you don’t claim it. If you do want to keep it, creating a free account adopts the one already trained on your content — nothing is crawled twice.