AI Answers API: ask ChatGPT, Perplexity, Gemini, Claude and Grok, with sources and a consensus

Ask ChatGPT, Perplexity, Gemini, Claude and Grok your questions in one run and get every answer with the sources it cites, searched live on the web from the country you pick. Hundreds of questions at once, as clean JSON for your app, sheet or agent.

Each question also gets a consensus report: the answer most of them support, what they agree and disagree on, claims only one makes, an agreement score and the sites more than one cited. Pick fast low-cost models (cents an answer) or each company's flagship.

What you get

Field	What it holds
Answer rows	prompt, engine, model, answer, citations (position, url, domain, kind), citationCount, citedDomains, country, webSearch, finishReason, aiCostUsd, seconds
Consensus rows	consensus, agreeOn, disagreeOn (each assistant's position), uniqueClaims, mostComplete, agreementScore, domainsCitedByMoreThanOne, enginesAnswered
SUMMARY	the models used, answers, failures, consensus reports, the AI usage reported, what was charged

What people use it for

• Compare the AIs: how each assistant answers the questions your customers ask, and which sites it trusts
• Research with sources: answers to many questions at once, each with links to check
• Fact-checking: where five assistants agree, and where one stands alone
• Apps and agents: multi-model answers through one API, without five API keys

How to use it

1. Add your questions, one per line.
2. Pick the assistants, the country to search from and the model tier.
3. Click Start and open the Answers table: one row per question and assistant, one consensus row per question.

Input

Field	Example	Notes
Questions	What are the best project management tools for a small team in 2026, and what does each cost?	Each goes to every assistant
AI assistants	ChatGPT, Perplexity, Gemini, Claude	Grok too
Models	fast	flagship: what each assistant's app runs
Live web search	on	Off: answers from what the model knows
Country	US	Gemini takes no location
Consensus report	on	Two or more assistants
Instructions	Answer in under 100 words	Sent with every question
Fixed models	{"chatgpt": "openai/gpt-6.1-sol"}	Any OpenRouter model

Output

Each result is one row in the run's dataset (download it as JSON, CSV, Excel, XML or HTML, or read it through the API). An example row (values shortened):

{
  "prompt": "What are the best project management tools for a small team in 2026, and what does each cost?",
  "engine": "ChatGPT",
  "model": "openai/gpt-5.4-mini",
  "modelTier": "fast",
  "webSearch": true,
  "country": "US",
  "answer": "Here's a practical 2026 shortlist for a small team… | Tool A | Teams that want an all-in-one workspace | Free; paid from $7/user/month (annual billing) ([tool-a.example](https://tool-a.example/pricing)) | …",
  "citations": [
    {
      "position": 1,
      "url": "https://tool-a.example/pricing",
      "domain": "tool-a.example",
      "kind": "other"
    },
    {
      "position": 2,
      "url": "https://tool-b.example/pricing",
      "domain": "tool-b.example",
      "kind": "other"
    }
  ],
  "citationCount": 7,
  "citedDomains": [
    "tool-a.example",
    "tool-b.example",
    "tool-c.example",
    "tool-d.example"
  ],
  "finishReason": "stop",
  "aiCostUsd": 0.0312,
  "seconds": 8.1,
  "status": "ok"
}

From runs on 17 October 2026 (one pricing question to all five assistants on the fast models; tool names and sites replaced): each answered with 0 to 20 sources, Perplexity in 5 seconds and Grok in 25; the consensus report gave an agreement score of 75 out of 100, named the most complete answer, listed the prices they disagreed on with each assistant's figure, and the claims only one of them made. A record named SUMMARY in the run's key-value store lists the models used and what was charged.

Pricing

• $0.004 per AI answer: one assistant's answer to one question, with its sources
• $0.004 per consensus report (one per question when two or more assistants answered; switch it off to skip it)
• Free: failed answers and failed reports
• The AI usage is billed by Apify's AI gateway at OpenRouter's rates. Measured 17 October 2026 on a hard pricing question, fast models: Perplexity Sonar $0.006, GPT-5.4 mini $0.02-0.03, Claude Haiku 5.5 $0.023, Grok 4.3 $0.049, Gemini 3.8 Flash $0.061 (Google charges per search, and it ran about four); the consensus report about $0.005. Flagship: GPT-6.1 Sol about $0.09 and Grok 4.7 $0.15-0.23 an answer (12 October); Apify charges free-plan accounts 10 times the AI usage

What typical jobs cost at the list price:

Job	Cost
1 question to 4 assistants with a consensus (fast)	$0.02 + about $0.12 AI
100 questions to Perplexity only	$0.40 + about $0.60 AI
10 questions to all five (fast)	$0.24 + about $1.70 AI

On Apify's Scale and Business plans the per-event prices are lower (Store discounts). Set Maximum cost per run and the Actor stops cleanly when it is reached: you never pay more than that.

Tips

• Perplexity alone is the cheapest way to get web answers with many sources.
• Use the flagship tier when you need the answers your customers see in each assistant's app.
• Shape every answer with Instructions: a length, a format (a table, a numbered list) or a role.
• Tracking whether the AIs name your brand? The AI Answer Tracker counts mentions and citations per brand.

Use it through the API

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("zaiq/ai-answers-api").call(run_input={"prompts": ["What are the best project management tools for a small team in 2026?"], "engines": ["chatgpt", "perplexity", "gemini", "claude"], "country": "US"})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(row)

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });
const run = await client.actor('zaiq/ai-answers-api').call({"prompts": ["What are the best project management tools for a small team in 2026?"], "engines": ["chatgpt", "perplexity", "gemini", "claude"], "country": "US"});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

HTTP (one call, results back)

curl -X POST 'https://api.apify.com/v2/acts/zaiq~ai-answers-api/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>' \
  -H 'Content-Type: application/json' \
  -d '{"prompts": ["What are the best project management tools for a small team in 2026?"], "engines": ["chatgpt", "perplexity", "gemini", "claude"], "country": "US"}'

Integrations and schedules

• Schedules: run it every hour, day or week from the Apify Console (Schedules) and keep a history of results.
• Make, Zapier, n8n, Google Sheets, Slack, Airtable: connect it through Apify's integrations, or send each finished run to a webhook.
• Datasets: results stay in the run's dataset; export them or read them by API whenever you need them.

Use it with AI agents (MCP)

Add Apify's MCP server (https://mcp.apify.com) to Claude, ChatGPT, Cursor, VS Code or any MCP client. The agent can find this Actor and run it for you: ask it for "ask ChatGPT, Perplexity and Claude this question and show me where they disagree". AI agents can also pay for runs directly through Apify's agentic payments.

Limits

• These are the assistants' official APIs with their own web search, not the consumer apps: the apps add memory, personalisation and their own instructions, so their answers can differ.
• Answers change from run to run and as the models are updated.
• An answer can hold at most 2,048 tokens (about 1,500 words), the limit of Apify's AI gateway.
• Gemini does not take a location; Perplexity always searches the web.

FAQ

Do I need API keys for OpenAI, Anthropic, Google or xAI?

No. Apify's AI gateway runs every model and bills its exact usage to your Apify account.

Is it the real ChatGPT?

It is OpenAI's model through OpenAI's API with OpenAI's own web search: the same model family as ChatGPT, not the chatgpt.com app.

Which models are used?

Each company's newest model in the tier you pick, chosen at the start of every run; the model is in every row. Fixed models pins any OpenRouter model.

Can I switch off web search?

Yes, and the assistants answer from what they know (Perplexity always searches).

Related Actors from Zaiq

• ai-answer-tracker: whether the AIs name and cite your brand
• bulk-ai-runner: one prompt on every row of your data
• ai-web-scraper: any page to JSON with a prompt
• llm-website-crawler: sites as Markdown for your own AI

Support

Questions, a site that changed, or a field you need? Open an issue on this Actor's Issues tab: we answer there, usually within a day.

---

Also found as: chatgpt api, perplexity api, ai answers api, llm api, compare ai answers, gemini api, claude api, grok api, ai search api, multi llm api.