AI share of voice · form builder · September 2026

When a buyer asks ChatGPT or Perplexity for form builder, who gets named?

We asked both assistants the twelve questions a buyer types when choosing a vendor — 24 live answers — and counted how often each vendor was named, and which websites the assistants read to decide. No vendor was involved. The numbers are below; the method is at the bottom.

Share of answers that name each vendor

Aliases, product names and old names are merged. Percentages are of all 24 answers. Vendors named only once are omitted.

The two assistants do not agree

ChatGPT and Perplexity build their answers from different pages, so the same vendor can be a default on one and near-invisible on the other. The largest gaps:

What the assistants read to decide

Every answer came with source links. These are the domains cited most across all 24 answers. Presence on the top few (fillout.com, forms.app, blog.hubspot.com, emailtooltester.com) determines the category far more than anything on a vendor's own site.

Method

The twelve prompts: Best form builder for startups · Compare the leading form builder tools. · Recommend a form builder for a team of 10. · What are the top form builder options in 2026? · What form builder do professionals actually recommend? · What is a good alternative to the most popular form builder? · What is the best form builder? · What is the most affordable form builder? · What should I look for when choosing a form builder? · Which form builder has the best reviews? · Which form builder should a small business use? · Who are the leading providers of form builder?

Each prompt was sent live to ChatGPT and Perplexity (ChatGPT as gpt-5 with web search on; Perplexity as sonar) via the DataForSEO LLM Responses API on 25 September 2026 — 24 answers. Every answer was checked for each vendor name using case-insensitive matching with aliases and old names merged; vendors were drawn from the answers themselves, so a vendor that no answer named does not appear. Source links inside each answer were counted once per domain per answer (up to eight domains per answer, so citation counts are lower bounds). Answers were truncated at 4,000 characters before scoring. Assistants are non-deterministic: a single vendor's share can move several points between runs. Total cost of the data on this page: US$0.46.

We re-run categories monthly. If a vendor on this list wants the per-question breakdown for their brand — which questions they are named for, which they are not, and which sources carried the answers that omit them — it is free; write to audits@agentbuilt.dev. Want it every month, with the month-over-month trend? Brand Watch does that for one brand at $39 a month.