MivoChat research · 5 November 2026
We scanned 158 SaaS websites: 91% leave visitor questions unanswered
Key findings
What the sample showed
158 of 200 attempted websites produced a report with at least three scanned pages.
- 5.1%
of scanned sites had a contradiction
- 1.3%
had likely outdated content
- 23%
had near-duplicate pages
- 91%
had at least one likely unanswered question
For editors
Press-friendly summary
We scanned 158 SaaS websites: 91% leave visitor questions unanswered. In a deterministic sample of 200 YC-listed B2B SaaS companies, MivoChat attempted 200 website scans and obtained 158 usable reports. Among scanned sites, 5.1% showed at least one contradictory claim, 1.3% had likely outdated content, and 23% had near-duplicate pages.
- 91% of scanned sites had at least one likely unanswered visitor question.
- Scans read up to 30 public pages per site and respected robots.txt.
- Results are anonymised aggregates, not a company ranking or individual audit.
Cite this research
Suggested citation: MivoChat. (2026, November 5). SaaS website content health 2026. https://mivochat.app/research/saas-website-content-health-2026
Contradictions
Where sites contradict themselves
Number of scanned sites with at least one contradiction in each category.
View as table
| Category | Sites | Share of scanned sites |
|---|---|---|
| dates | 3 sites | 1.9% |
| contact | 1 site | 0.6% |
| limits | 1 site | 0.6% |
| pricing | 1 site | 0.6% |
| security | 1 site | 0.6% |
| other | 2 sites | 1.3% |
Unanswered questions
What visitors ask that sites don't answer
Number of scanned sites with one or more likely unanswered questions in each topic.
View as table
| Category | Sites | Share of scanned sites |
|---|---|---|
| trial | 60 sites | 38% |
| limits | 53 sites | 34% |
| pricing | 48 sites | 30% |
| refunds/cancellation | 29 sites | 18% |
| support | 28 sites | 18% |
| integrations | 13 sites | 8.2% |
| security & compliance | 8 sites | 5.1% |
| data export | 1 site | 0.6% |
| other | 81 sites | 51% |
Method
How we built the sample
Source: the public YC company directory through the yc-oss API. We retained active B2B companies with a website, team size from 5 to 200, and nonprofit=false; website origins were normalised and deduplicated. A deterministic seed (mivochat-2026) selected 200 companies.
Of 200 sampled sites, 200 were attempted; 158 were scanned successfully and 42 failed (including 0 analysis failures). The crawler respected robots.txt, identified itself with the MivoChatBot user agent, and capped each scan at 30 pages. Scan date: 5 November 2026. Directory fetch date: 5 November 2026. The analysis used MivoChat's answer model and the scans used an embedding model. 14 included reports were marked partial.
The report is rendered from the aggregate JSON. It contains no company domains or text examples.
Limitations
How to read these results
- The sample is biased toward YC-listed B2B companies and does not represent all SaaS websites.
- Each scan read at most 30 pages, so deeper sites may have issues outside the sample.
- Model-assisted detection can miss issues or misjudge correct content.
- This is a single snapshot taken on 5 November 2026.
- Sites that failed crawling or produced fewer than three pages were excluded from scanned-site percentages.
- 14 included reports were marked partial, so some scans may not cover all reachable pages.
- No company is named and raw per-site reports are not published.
Questions
Research FAQs
What did this study measure?
We scanned 158 sampled SaaS websites for contradictory claims, likely outdated details, near-duplicate pages and likely visitor questions that the site did not answer.
How were the websites selected?
We sampled 200 eligible companies from the public YC company directory using a deterministic seed, then attempted to scan them.
Does this identify individual companies?
No. We publish aggregate counts only. Individual company names, domains, page text and scan reports are not included.
Are these findings definitive?
No. Model-assisted detection can miss issues or flag content incorrectly. Treat the aggregate as a snapshot and not a ranking or audit of any company.
Can I check my own site?
Yes. The free website content checker reviews up to 30 public pages and gives you findings to verify.
Data & privacy
Anonymised aggregate only
This report does not name individual companies. We publish aggregate counts only; raw per-site data, domains, names and page snippets are not published.
Make your site easier to trust
Find conflicting, stale and repeated details across your public pages.