Run the audit on the prospect's own domain, pick one finding they can verify in thirty seconds, and open with that. Here is how the website audit node works, how to use it, and where it stops helping.
Lumina7 min read
Run the website audit on the prospect's own domain before the call, pick the one finding they can verify in thirty seconds, and open with that instead of a pitch about AI. The audit works because the input is theirs. They know their site, so a finding about it is either true or it is not, and when it is true you have earned the next five minutes.
That is the short version. Below is what the audit node does on a Lumina Sales demo, how to turn its output into a conversation, and where it stops being useful.
What does the website audit node do?
It is one of the six nodes you can put on a demo page, and the only one that works on something the prospect already owns. You type their business name and website when you build the demo. The audit fetches that live domain and reports findings against it, such as the "6 issues found" you can see in the build preview on our homepage. We walk through all six nodes in what each node on the demo site actually does, so here we stay on the audit.
Two facts shape how you use it. The audit runs against the real site, not a canned example, so it can be wrong in the ways real fetches are wrong: a site that is down, slow, or blocking automated requests. And it counts against a monthly allowance of 50 audits, where one audit of one domain is one audit. Both matter below.
Why audit a prospect's site before you sell anything?
It swaps a claim for evidence. Most cold outreach about AI is a claim: "we can help you capture more leads." The prospect has heard it from six vendors this year. A finding about their own website is different, since they can check it right now.
Insites, a company that sells website audit software to agencies, lists seven reasons to audit a prospect's site first. The ones that carry over to AI services are qualifying the prospect before you spend time on them, personalizing outreach, and having "hard facts" such as broken links that the owner may not know about. Insites also says that when you cold message, you might hear back from 20% of people and only 5% book a call after the first message. That is a vendor's framing of its own market, not a benchmark, but the direction is the part to keep: a specific, checkable observation gets a reply more often than a generic one.
One of Insites' customer stories claims a jump from 1 in 80 to 1 in 25 outreach conversions after an agency started running audits live on calls. The page is on the vendor's own site and gives no methodology, so treat it as a story and not a forecast. We would not promise you that number. We would tell you the mechanism is plausible.
How do you turn audit findings into a conversation?
Run it yourself first. Always. Open the published demo page, let the audit finish, and read each finding against the actual site in another tab. If a finding is wrong, you want to find that out alone, not while the owner is watching.
Then use a short sequence.
Pick one finding, not all of them
An audit that returns six issues gives you six ways to overwhelm someone. Choose the single finding that is easiest to verify and most expensive to ignore. A contact page that does not load on a phone beats a vague note about metadata. The owner will confirm it with one tap and then ask you what else you found, which is the question you want.
Tie the finding to the call they are missing
Your product is not a website fix. You sell a system that answers calls, texts back, books jobs, and follows up. The bridge is simple: the site is where a lot of their after-hours enquiries arrive, and the audit shows how well it handles them. A slow page or a buried phone number is a missed-lead story in miniature. Then you move from the site to the receptionist node, and the demo does the rest.
Use Google's own numbers when the finding is about speed
If a finding is about slow loading, check it against the published thresholds before you repeat it. Google's web.dev lists the Core Web Vitals "good" thresholds as a largest contentful paint of 2,500 milliseconds or less, an interaction to next paint of 200 milliseconds or less, and a cumulative layout shift of 0.1 or less, assessed at the 75th percentile of page views. Quoting a number the owner can look up beats saying "your site feels slow." I am not claiming our audit reports those exact metrics. If a finding mentions speed, though, you should be able to point to a standard.
Let the dashboard tell you whether they ran it
After you send the link, the dashboard shows whether the prospect opened it and whether they ran the audit or talked to the receptionist. We cover what that activity does and does not tell you in what you see after a prospect opens your AI demo. A prospect who audited their own site has told you something. They cared enough to check.
What are the limits of using the audit this way?
There are five, and you should know them before you build outreach around it.
First, an audit proves a website has problems. It does not prove the owner needs an AI system. Plenty of businesses with messy sites have a full front desk and no pain. Use the audit to open a conversation, then qualify the old-fashioned way: how many calls go unanswered, what a missed job costs, who answers after five.
Second, it needs a website. Many of the smallest local businesses have a thin page, a Facebook page, or nothing. If your list is full of those, the audit has little to say and you are better off leading with the missed-call recovery node or the receptionist.
Third, the fetch can fail. A site that is down, rate limits requests, or sits behind a bot check can produce an empty or odd result. That is why you open the page before you send it.
Fourth, the allowance is 50 audits a month. If you plan to audit a hundred domains for research, you will hit it. Spend audits on prospects who are real, not on a spreadsheet of cold names, and save the bulk research for tools built for that. The plan limits are listed next to the price for this reason.
Fifth, do not oversell it. Telling someone their site has a flaw they could have found themselves is not a revelation. It is a credibility check. What it buys you is a reason for the prospect to believe your next sentence, the one about calls.
Who should not lean on the audit?
If you sell to businesses that live on phone calls and referrals, such as towing, locksmiths, or emergency plumbers, the owner may not care about their site at all. They care about the phone ringing at 2am. For those trades, put the voice receptionist first and use the audit as a back-pocket node. Our demo call playbook covers how to choose the opening node by what the prospect is most skeptical about.
FAQ
Does the audit tell the prospect their site is bad?
It reports findings, and how you frame them is up to you. Start with the one you can verify, say what it likely costs them, and move on. Do not read the whole list aloud.
Is one audit per domain, or per click?
One audit of one domain counts as one audit against your monthly allowance of 50. Run it yourself before you send the link so you know what the audit says about that domain before the prospect does.
Can I run the audit on a prospect's site without sending a demo?
The audit is a node on the demo page, so you build a demo for the prospect and publish it. That also means the audit and the receptionist sit together on one link, which is usually what you want for the follow-up.
What if the audit result looks wrong?
Do not send the link. Check the site in a browser, and check whether it blocks automated requests. If the finding is wrong, drop that node from the page or lead with a different one. A false finding costs you more trust than no finding.
What to do next
Pick one prospect with a real website, build their demo in a few minutes from the live demo setup, and run the audit on their domain. Write down the one finding you could verify in thirty seconds. If you do not have a workspace yet, start one for $49 a month and use that finding as your opening line on the next call.
