Vapi already has the largest library a voice-agent buyer could search for: 291 templates from dental offices to freight claims. But every one we sampled tells search engines and AI assistants it is the homepage, with the same title and description, and some still show older models. The pages exist; they just do not say what they are.
First action
Change the template page component so the title and description come from the template's own heading and summary.
How to measure it
Sign-ups from /custom-agents pages
- Now
- –
- Target
- +50%
- By
- 90 days
Steps
Title from data
Use each template's H1 and first sentence for the title and description.
Frontend engineer 1 day
Add the context
Industry, call type (inbound, outbound, scheduling) and "Fork in Vapi" in the description.
Product marketing 1 day
Refresh the defaults
Swap GPT-4o and Nova 2 defaults for the current recommended stack.
Senior Product Manager 1 day
Fix odd slugs
Redirect the timestamped slug to a clean one.
Frontend engineer 1 hour
Cross-link
Each template links to the closest customer story and docs page.
Product marketing 2 days
Measure
Track impressions, clicks and forks per template in PostHog.
Frontend engineer 2 hours
Why it works
Generate a unique title and description for each template from its own heading and use case ("AI voice agent for dental offices: booking and rescheduling | Vapi"), add the industry and the call type, refresh the default model, transcriber and voice to current ones, clean the timestamped slug with a redirect, and link each template to the matching customer story and the docs.
Template searches ("AI receptionist for dental office", "voice agent for freight claims") are high-intent: the searcher wants to build that exact agent. Fixing 291 titles is a one-template change in code, and the PLG inbound the Enterprise AE inherits starts with someone forking an agent.
Ready to use
ChecklistTemplate page fields
[ ] Title: "AI voice agent for <use case> | Vapi" [ ] Description: what the agent does, the industry, "fork and customise in minutes" [ ] H1 matches the title [ ] Current default model, transcriber and voice [ ] Link to the matching customer story [ ] Link to the docs for the tools it calls [ ] Clean slug (no timestamps)
Risks and how to handle them
Hundreds of thin pages look like spam
Keep the system prompt and configuration on each page; it is real, unique content.
Model defaults age again
Pull defaults from the same config the dashboard uses.
Evidence
- Your websiteSitemap (October 18, 2026): 546 URLs, of which 291 are /custom-agents/ templates and 217 are blog posts. Every template page we sampled (20 of them) uses the homepage title "Vapi - Build Advanced Voice AI Agents" and the homepage description; one slug ends in "-20250522-114257". The dental office template still specifies OpenAI GPT-4o and Deepgram Nova 2.Source
- HiringEnterprise Account Executive ($150K–$175K base): Vapi has "repeatedly proven our ability to sell solutions into the enterprise" with Amazon, Zillow and Uber and is "now primed to expand into new use cases and verticals". Reps get inbound "from our PLG motion, where prospects come in and build agents on our platform and are highly activated before they reach the sales team", outbound with an SDR, and must close a first deal "at an ACV of $240k+" in 90 days with 3x coverage.Source
- Your websiteHomepage: "Voice agents for builders"; "1 Billion calls supported", "99.9% uptime for enterprise clients", "2.5M+ agents launched", "1M+ developers", "<500ms average latency"; "Everything a Fortune 100 needs to deploy at scale"; an Amazon Ring quote ("zero to production in two weeks, and 100% of our inbound volume now runs through Vapi"); a VapiCon banner for November 11–12.Source