SERP APIs are built for agents and govern their own contracts least

SERP APIs are built for agents and govern their own contracts least

The SERP tag on apis.io is small — 31 providers and 36 APIs — and it is one of the most agent-ready corners of the catalog. Its members average a Kin Score of 51.2 against 20.4 for the rest of the catalog, and not one of the 31 is human-only, against 53.8% of everyone else. These are the companies that turn a search engine results page into JSON, and in a year when every agent needs to look something up, they have been building for that reader longer than most.

The five the tag page ranks first. SerpApi leads at 75.0, exemplar, carried by operational transparency and access clarity at 92.1 each, with 10 API pages, 2 MCP servers and an agentic-access profile; it is the reference product for scraped results across Google, Bing and Baidu. DataForSEO is second at 73.6, exemplar, the strongest developer ergonomics in the five at 90.5, and the widest agent surface: delegated identity, dynamic client registration and protected resource metadata all lit. Zyte is third at 68.4, exemplar, the web-scraping platform whose access clarity of 93.4 is the best in the cohort. ValueSERP, Traject Data’s results API, is fourth at 66.5, exemplar. Chocodata closes the five at 66.2, strong, a structured-data extraction service with an MCP server. Note that the tag spans two kinds of company — dedicated results APIs and general scraping platforms — and the five are split between them.

The cohort against the market. The median member scores 55.1 against 13.5 for the rest of the catalog. On the Kin Score ladder, 12.9% of SERP providers are exemplar and 38.7% strong, against 2.0% and 4.3% elsewhere; there are no minimal providers at all, where 40.8% of the rest of the catalog sits. Every scored member publishes a machine-readable contract — spec presence is 100.0% against 32.7% — and 80.6% put examples in their OpenAPI against 15.6%. This is a category that sells API calls and nothing else, and it documents like it.

Agent readiness. The mean is 40.1 against 12.0. Six members are agent-native (19.4% against 2.0%) and twenty agent-ready (64.5% against 14.4%). The gaps are wide almost everywhere: auth clarity is lit for 100.0% of the tag against 38.3%, error semantics for 80.6% against 22.9%, rate-limit signals for 83.9% against 30.6%, and 58.1% run an MCP server by the tag’s dimension count against 16.7%. The dark spots are the ones you would expect from a read-only product: dry-run mode at 0.0% and reversibility at 12.9% matter little when nearly every call is a lookup. One pairing is worth a sentence. DataForSEO has the highest Agent Readiness score in the five at 61.3 and is recorded agent-ready, while SerpApi at 41.9 and ValueSERP at 40.5 are agent-native. The band is gated on specific dimensions, not read off the score, and that is the rubric working as designed.

The observation. The weak facet is not the contract itself but the governance around it. Contract governance is the lowest facet for every one of the five — 41.7 at DataForSEO, 35.6 at SerpApi, 28.0 at ValueSERP and 18.2 at both Zyte and Chocodata — while contract quality runs from 49.0 to 72.3. SERP providers publish specs an agent can read on the first try, and say much less about how those specs are versioned, changed and retired. For a category whose output shape is dictated by a search engine that changes its page whenever it likes, that is the promise an agent most needs and the one this tag makes least.

← Our deprecation-policy feature lists five providers, and three matched on storage expiry
Ably's Rules API routes messages to 14 destinations, and its catalog record points at the wrong host →