When I reviewed the remodeling website generated by Lovable, the photographs were appealing and the navigation was clear. The text immediately concerned me.
The service descriptions were vague. They provided little specific information about the work offered or why a prospective customer should choose the company. Important credibility information, including licensing, insurance, and credentials, blended into the surrounding text.
My earlier experience in website conversion shaped that reaction. I look at whether a website gives people the information they need to make a decision. Does the company offer what they need? What supports its claims? Do visitors have enough confidence to contact it?
My approach was influenced by Bryan and Jeffrey Eisenberg's work on website conversion and Persuasion Architecture: considering how customers make decisions and what information they need to move forward. That perspective shaped my review of the AI-generated page.
From that perspective, the AI-generated website needed substantial human editing before I would consider it ready for a real business.
Looking beyond the appearance
I compared the main page of the Lovable prototype with the main page of J.S. Brown & Co., an established remodeling company with a non-AI-generated website.
The difference in written content stood out. J.S. Brown & Co. provided specific information about their services and experience. Recognition and accreditation logos made its credentials noticeable. Its project photographs appeared connected to the company's work.
Lovable produced an attractive layout, but its brief descriptions left important questions unanswered. There was room on the page to explain the services, demonstrate the company's capabilities, and give visitors reasons to choose it.
Credentials were present, but their presentation did little to draw attention to them. Contact prompts also needed greater visibility throughout the long page, with clearly labeled alternatives to calling.
These were findings from my expert review. I wanted to examine prospective customers' impressions separately.
What 54 participants thought
I conducted a survey in which 54 participants viewed a screenshot of J.S. Brown & Co.'s main page. Participants were recruited through LinkedIn and Facebook and were not paid.
They rated their trust in the company, confidence in its qualifications, perceived professionalism, and confidence that it would deliver quality work. They also rated their likelihood of contacting the company and requesting a quote.
Their impressions were consistently positive:
Participants also described what influenced their impressions. Licensing, insurance, and credentials were the most frequently mentioned theme, followed by visual appearance and professionalism, then customer reviews and reputation.
These comments identified information participants noticed and considered relevant. They did not verify which page elements caused the positive ratings.
Participants did not evaluate Lovable in this survey, so the findings cannot establish which website they would trust more.
Trust and willingness to take the next step
I used R to explore relationships within the survey responses.
Higher trust ratings were strongly associated with greater stated likelihood of contacting the company (Spearman’s p = .712) and requesting a quote (p = .678). Both relationships remained statistically significant after Holm adjustment for the two correlations (adjusted p <.001).
People who expressed greater trust also tended to express greater willingness to take the next step.
That connection is relevant to a website conversion review. It supports examining trust alongside contact intentions. It does not show that trust caused those intentions, or that participants would actually contact the company.
Experience and hiring plans added context
Of the 54 participants, 38 had previously hired a contractor and 16 had not.
Among those with previous contractor experience, 35 of 38 (92.1%) were very or extremely likely to request a quote. Among those without that experience, 10 of 16 (62.5%) selected those responses.
The difference in quote-request ratings remained statistically significant after Holm adjustment across six rating comparisons (adjusted p = .027). The other five comparisons did not remain statistically significant.
Future hiring plans were also associated with one specific rating: likelihood of contacting the company.
Among participants expecting to hire a contractor within 12 months, 35 of 37 (94.6%) were very or extremely likely to contact the company. That compared with 7 of 11 (63.6%) who were unsure and 4 of 6 (66.7%) who did not expect to hire.
The overall difference remained statistically significant after adjustment (adjusted p = .014). Follow-up comparisons found higher contact-likelihood ratings among those planning to hire than among those who were unsure (adjusted p = .009). The other pairwise comparisons were not statistically significant.
These findings suggest that participants' circumstances matter when interpreting their stated willingness to contact a business. Someone's interest in making an inquiry may reflect both their impression of the company and whether they anticipate needing its services.
What the AI-generated page still needed
The survey and expert review served different purposes. The survey examined impressions of the real company's page. My review examined how each page supported a visitor's decision.
Together, they sharpened the questions I would ask before using an AI-generated website for a real business:
Does the text provide useful details?
Is credibility information easy to notice?
Do photographs provide evidence
Is contacting the company straightforward?
AI-generated claims, credentials, testimonials, and images also need verification before publication. A convincing presentation cannot substitute for accurate business information.
What this study can-and cannot-tell us
This was an exploratory study with 54 self-selected participants. They viewed a screenshot rather than navigating the website. Some comparison groups were small, particularly participants who did not expect to hire a contractor.
The findings describe this sample's impressions and stated intentions. They do not measure actual inquiries, quote requests, or conversion rates. The website comparison reflects my assessment as one researcher, and participants did not evaluate the AI-generated prototype.
Those limits matter. They also leave a clear next step: revise the AI-generated content, test it with prospective customers, and evaluate whether visitors can understand the services, assess credibility, and contact the company.
The work after generation
Lovable produced a visually appealing starting point. My conversion experience helped me recognize substantial weaknesses in what the page actually told prospective customers.
The remaining work required judgment about the business, its customers, and the information people need before entrusting someone with their home.
That is the lesson I took from this study: generating a website does not complete the work of making it useful, credible, and ready to publish.
Email: theresaw@columbus.rr.com
LinkedIn: theresa-wilkinson