Nine SERP APIs, one response body: what actually differs
Field names, block coverage, vertical count and metering policy — in roughly that order of how much trouble they cause.
30 Jul 2026
We checked eight competing Google search APIs against their own documentation in August 2026. The conclusion that surprised us: the response bodies are more alike than the marketing suggests, and the differences that cost you time are not the ones anybody advertises.
Field naming splits the market in two
Serper, Zenserp, HasData and ourselves use camelCase. SerpApi, SearchApi and DataForSEO use snake_case. That single choice decides whether switching providers is a base-URL change or a parser rewrite, and it is the first thing to check.
Block coverage varies more than vertical count
Vertical count is the headline number — twenty against thirteen against two — but most integrations call one vertical. What bites is whether the web search response actually carries People Also Ask, a knowledge panel and sitelinks. Several providers documented fewer of these than their positioning implies.
Metering is the undiscussed difference
SerpApi does not bill errored or cached searches. We bill every request to a billed path, and say so on the pricing page rather than in a support thread. At a thousand calls a month this is noise. At a million it is the invoice.
Almost nobody returns the AI Overview
Of the eight providers we checked, one besides us documented an AI Overview field. Given that Google now writes the answer itself on around half of the queries we measure, that is a strange place for a market to be.
The full table, with dates and sources: how we compare.
Run the same measurements yourself
Free tier, no card. Every figure above came from one endpoint on a weekly schedule.