# Check that a citation or claim exists

> Check that a citation or claim exists: 4 companies fully do it, 14 with a limit, 68 carry the tag. Every listing in Web search APIs with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/search.verify-citation/
- Markdown: https://www.anchorterminal.com/niches/search.verify-citation/index.md (~1,700 tokens)
- Slim: https://www.anchorterminal.com/niches/search.verify-citation/index.min.md (~430 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/search.verify-citation/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

Four or more companies' docs show a full fit.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 4 | 14 | 68 | 6 | 2 |

## What counts as done

For each item, a live result confirming or denying it, at a sustained rate of at least 5 a second without a rate-limit stall, and no model-written answer in place of the source.

How agents ask for it: do these sources exist; verify this quote; is this URL real and does it say that

## Does the job

The docs show it meets the test above, and nothing we read says otherwise.

- **Web Search Tool on Amazon Bedrock AgentCore (Amazon Web Services, Inc. (regional AWS entities by account location))**: Raw results with URL and title, no model-written answer, default quota 10 Web Search requests a second per account and Region, adjustable, with queries kept inside AWS.
- **Exa API + MCP (Exa Labs Inc.)**: Raw results at 10 QPS for /search and 100 QPS for /contents on the free tier, 402 on exhausted credit, no model-written answer in the search path.
- **Linkup (Linkup Technologies SAS)**: Search and fetch return raw sources at 10 queries a second per organisation, with zero data retention on request.
- **Parallel Search and Task APIs (Parallel Web Systems Inc.)**: Search and Extract each allow 600 requests a minute and return raw excerpts or page markdown with URLs.

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **Brave Search API + MCP (Brave Software, Inc.)**: Raw results, but the per-second limit is set by plan and a 429 carries X-RateLimit-Reset rather than Retry-After; 500-user deployments report constant 429s.
- **DeepSeek Search and Fetch API (Hangzhou DeepSeek Artificial Intelligence Co., Ltd.)**: Raw results and pages, but 5 requests a second per account across keys is the map's blocker, and a citation-checking user will not use it for lack of zero data retention.
- **Fetch (MCP reference server) (Model Context Protocol, a Series of LF Projects, LLC)**: Returns the raw page with no provider rate limit, but a robots.txt disallow, a 401, 403 or 5xx on robots.txt, and JavaScript-only pages all stop a check.
- **Liner Search MCP (Liner)**: Raw web results with no per-second limit (fair use), but no fetch, and scholar and agent calls are capped at 2 a second.
- **Mojeek Web Search API (Mojeek Limited)**: Raw results at 5 queries a second on Startup and 10 on Business, but no fetch and an over-limit error comes as text in a 200.
- **NewsCatcher (NewsCatcher, Inc.)**: Can confirm a news article exists, but requests a second are set per plan and read from GET /subscription rather than published.
- **Perplexity Search API (Perplexity AI, Inc.)**: Raw results and a fetch tool, but QPS rises with cumulative spend from 1 on Tier 0, a user checking 100 citations a job needs higher limits, and ZDR needs a written agreement.
- **SearchAPI.io (SearchApi, LLC)**: Raw SERP results, but no per-second rate is documented and monthly plans pace hourly.
- **SerpApi (SerpApi, LLC)**: Raw SERP results and free cached repeats, but plans are hourly paced and a 429 means either the hourly limit or an empty balance.
- **SerpKite (Anastasia Klimova, trading as SerpKite)**: Raw results and page fetches with distinct 429 rate_limited and 503 codes, but no requests-a-second figure appears in the spec.
- **SpecStream (SpecStream SAS)**: A fact answer is returned only when two sites print the value or a vote rates it likely; the per-minute call limit is not published.
- **Tavily API + MCP (AlphaAI Technologies Inc. dba Tavily)**: Raw results and extracts, but the free plan allows 100 requests a minute, under 5 a second, and no zero data retention option was found.
- **Valyu (Valyu.Network Ltd)**: Raw results with URLs, but 5 requests a second on pay-as-you-go (10 on the $29 plan), billed per result, with 206 for partial failures.
- **You.com APIs (SuSea, Inc., d/b/a you.com)**: Raw results and contents, but rate limits depend on an unpublished tier and ZDR covers only Web Search and Answer on enterprise agreements.

## Couldn't check

The docs were closed to us, or silent. These carry no mark either way.

- **Cloudflare Web Search API (Cloudflare, Inc.)**: No rate limit for the endpoint on the pages read, so the 5 a second criterion cannot be checked from docs.
- **Kagi Search API (Kagi Inc.)**: No rate numbers are published and a 429 covers both rate limiting and an exhausted usage limit, so 5 a second cannot be read from the docs.
- **Serper (Serper)**: No rate limit is published anywhere readable; the playground needs JavaScript.

- not this job: Arcmira (Arcmira, Inc.): Can confirm only that something was said in an indexed video or podcast, not that a web URL exists.
- not this job: Grounding with Bing Search (Microsoft Corporation): The provider returns an answer rather than the raw source, the blocker the map names for Bing Grounding.
- not this job: Grep AI (Parcha Labs, Inc.): A run returns a report, not a live result per URL at 5 a second.
- not this job: SERPHouse (Patoliya Infotech LLP): 60 requests a minute is one a second, below the 5 a second the use case needs.
- not this job: TinyFish Search (Tiny Fish, Inc.): 30 requests a minute per person is half a request a second, far below the 5 a second the use case needs.

## What stops agents finishing

- per-account rate limit of 5 a second shared across keys (DeepSeek, TinyFish 30 a minute)
- provider returns an answer rather than the raw source (Bing Grounding)

## Around it

Before: search.deep-research. 

## Where this comes from

Weight 6 of 100 in Web search APIs: 2 of 44 accounts (one detailed, one vendor-aligned); weighted above its count because it is the step that makes the other research jobs trustworthy

As data: https://www.anchorterminal.com/api/v1/use-cases/search.verify-citation.json, or the MCP tool get_use_case with id search.verify-citation. Draft, read 2026-11-12 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
