How accurate is the token count?

It's an estimate at ~4 characters per token. Exact counts depend on the model's tokenizer. Use it to compare formats rather than as an exact bill.

Where do the LLM prices come from?

They are editable assumptions, not a price list. Enter your model's price per million input tokens to get your own cost per 1,000 calls.

Does format=markdown cost more credits?

No. json, compact and markdown all cost the same credits. You only save tokens.

Can I paste a response from another SERP API?

Yes. Any JSON works. The compact and Markdown conversions look for SerpKite v1 keys (results, answer_box, people_also_ask); other shapes are counted as-is.

Free tool

SERP token counter

Paste a search API response (or any text), or run a live query, and compare the token cost of each output format before it hits your context window.

Format	Tokens	vs largest	Small modelper 1k calls	Mid-size modelper 1k calls	Frontier modelper 1k calls
JSON, pretty-printed What most SERP APIs return	668		$0.10	$0.668	$2.00
JSON, minified Whitespace removed	534	−20%	$0.08	$0.534	$1.60
format=compact Titles, URLs and snippets	228	−66%	$0.034	$0.228	$0.684
format=markdown LLM-ready prose	231	−65%	$0.035	$0.231	$0.693

Why output format matters for agents

A full SERP JSON is mostly structure: keys, quotes, positions, sitelinks, tracking fields. An agent needs titles, URLs and snippets. Every extra token costs money on each call and crowds the context window.

SerpKite can return format=compact (JSON with only what agents need) or format=markdown (prose an LLM reads directly). Both typically cut 60–80% of the tokens of a full SERP, and fields= trims further.

How the estimate works

We use about 4 characters per token, the same heuristic the API uses for the X-Tokens-Estimate header. Real tokenizers differ by model, so treat the numbers as a comparison between formats rather than an exact bill. The compact and Markdown rows are built in your browser from the JSON you paste.

The cost columns multiply tokens by the input price you enter. The defaults are illustrative, so replace them with your model's price.

Do it with the API

Cost: 1 credit, any format. Failed and empty results are free.

POST /v1/search

cURL

Python

TypeScript

Go

fetch

PHP

curl https://api.serpkite.com/v1/search \
  -H "Authorization: Bearer $SERPKITE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"q":"best espresso machine 2026","format":"markdown"}'

FAQ