Google AI Mode API
Google's generated answer for any query, as typed text blocks with the sources it cited — one POST returning structured JSON. The surface where citation is replacing ranking, made measurable.
50 free credits on signup. No card required.
{ "search_parameters": { "query": "Best AI Model 2026", "hl": "en", "gl": "us", "device": "desktop", "google_domain": "google.com" }, "text_blocks": [ { "type": "heading", "snippet": "Elite Models Stackup by Discipline", "level": 3 }, { "type": "paragraph", "snippet": "Use Case / Domain Best Model Key Strength Complex Reasoning & Planning…" } ], "references": [ { "title": "AI Comparison 2026: The Best AI Models in a Data Overview", "link": "https://www.punku.ai/blog/..." } ], "credits_used": 1, "credits_remaining": 4745}Live call against the real endpoint — 3 free runs, then it's your own key.
What is the Scavio Google AI Mode API?
The Scavio Google AI Mode API is a REST API that returns Google's AI-generated answer for a query as structured JSON. You send a query with an optional country and language to a single POST endpoint with your API key, and get back the synthesised response as typed text blocks plus the list of references Google cited.
- 1
- endpoint, live in production
- 1
- credit per call
- 0
- official APIs for this
One endpoint, the answer and its sources
What Google says, and who it credited for saying it.
AI Mode
POST /api/v2/google/ai-modeGoogle's generated answer for a query, broken into typed text blocks, with the list of sources it cited — the surface that is quietly replacing ten blue links.
The Google AI Mode endpoint
A POST that returns JSON and takes the same API key as every other platform. 1 credit per successful call.
| Endpoint | Returns | Credits |
|---|---|---|
AI Mode/api/v2/google/ai-mode | Synthesised answer as text_blocks, plus the references it cited | 1 |
50 free credits on signup covers 50 AI Mode queries. See plans
The answer, and whether it cited you
Read the blocks, then check the references for your domain. The response on the right is a real capture — note the flat shape, with no data wrapper.
from scavio import ScavioClient
client = ScavioClient(api_key="sk_live_your_key")
res = client.google.ai_mode(query="Best AI Model 2026")
for b in res["text_blocks"]:
print(b["type"], b["snippet"][:80])
# Who Google cited - are you in the list?
for r in res["references"]:
print(r["link"]){
"search_parameters": {
"query": "Best AI Model 2026",
"hl": "en",
"gl": "us",
"device": "desktop",
"google_domain": "google.com"
},
"text_blocks": [
{
"type": "heading",
"snippet": "Elite Models Stackup by Discipline",
"level": 3
},
{
"type": "paragraph",
"snippet": "Use Case / Domain Best Model Key Strength Complex Reasoning & Planning…"
}
],
"references": [
{
"title": "AI Comparison 2026: The Best AI Models in a Data Overview",
"link": "https://www.punku.ai/blog/..."
}
],
"credits_used": 1,
"credits_remaining": 4745
}Why you cannot just check this by hand
AI Mode has no official API, and its output moves. One manual look tells you almost nothing.
Checking manually
- No public first-party API exists for AI Mode
- Output varies by market, device and session, so one look is not a measurement
- Personalisation and location make your browser an unreliable instrument
- Tracking a keyword set by hand does not scale past a handful of terms
Scavio Google AI Mode API
- A documented endpoint with a stable, typed response shape
- Market held constant per call with explicit gl and hl
- References returned as a list, so citation tracking is a set operation
- Same key and JSON envelope as 30 other Scavio platforms
What developers build with it
AEO and citation tracking
The question that replaced rank tracking: does Google's generated answer cite you? Run your commercial queries on a schedule and record which domains appear in references over time.
ai-modeCompetitive share of answer
Count how often each competitor is cited across a keyword set. That share is the new page-one share, and it moves independently of classic rankings.
ai-modeContent gap analysis
Read the text blocks Google synthesises for a topic and compare them against your own coverage. What it asserts and you do not address is the brief.
ai-modeAnswer-quality monitoring
Track whether Google's answer about your product or category is accurate, and catch a confidently wrong claim before your sales team hears it from a prospect.
ai-modeAgent grounding
Point an agent at the MCP server and AI Mode becomes a callable tool — a synthesised answer with its sources attached, rather than ten links to read.
MCP + RESTHow to get Google AI Mode answers as JSON
- 1
Get an API key
Sign up for a Scavio account and copy your key from the dashboard. You get 50 credits free on signup, with no card required.
- 2
POST a query
Send {"query": "Best AI Model 2026"} to https://api.scavio.dev/api/v2/google/ai-mode with an Authorization: Bearer header. Set gl and hl to choose the market.
- 3
Read blocks and references
Google v2 responses are flat — text_blocks and references sit at the top level, not under data. Blocks are typed (heading, paragraph) so you can render or parse them; references are the cited sources.
Google AI Mode API FAQ
What is the Scavio Google AI Mode API?
+
It is a REST API that returns Google's AI-generated answer for a query as structured JSON. One POST endpoint returns the synthesised response as typed text blocks plus the references Google cited. You authenticate with a Scavio API key.
Who is this endpoint for?
+
SEO and AEO teams tracking whether they are cited; developers embedding synthesised answers in a product; AI agent builders grounding responses with sources attached; content strategists finding gaps; and competitive analysts measuring share of answer. Same API for all of them.
What is the difference between AI Mode and AI Overview?
+
AI Overview is the generated block that appears above classic results on a normal search — it comes back inside the standard Search API response. AI Mode is Google's dedicated conversational answer surface, and this endpoint targets it directly.
Why does this matter for SEO?
+
When Google answers a query itself, the click that used to reach a page one ranking may never happen. What replaces rank as the thing worth tracking is citation: whether your domain appears in the references behind that answer. This endpoint measures exactly that.
Does Google offer an API for this?
+
No. AI Mode has no public first-party API, and its output varies by market and device, so a single manual check tells you little. A programmatic endpoint is the only practical way to track it across a keyword set over time.
How much does an AI Mode request cost?
+
1 credit per call. New accounts get 50 free credits on signup, one time, with no card required. Paid plans start at $30 per month for 7,000 credits.
What does the response look like?
+
Google v2 responses are flat rather than wrapped: search_parameters, text_blocks and references sit at the top level, alongside credits_used and credits_remaining. Each block carries a type such as heading or paragraph, a snippet, and a level for headings.
Are the answers stable between calls?
+
Not exactly. Generated answers vary by market, device and over time, which is why tracking them is a time series rather than a one-off lookup. Set gl and hl explicitly so at least the market is held constant across runs.
Which languages and tools are supported?
+
Official Python (pip install scavio) and JavaScript/TypeScript (npm i scavio) SDKs, a remote MCP server at mcp.scavio.dev for AI agents, and an n8n community node. Any language can call the REST endpoint directly with an HTTP client.
Explore more of Scavio
Google Search API
The full SERP, including AI Overview.
Google Trends API
Interest over time and trending now.
Google News API
Headlines, sources and dates as JSON.
Extract
Any URL as Markdown, text or raw HTML.
MCP server
Every platform as a callable tool for agents.
Pricing
Credit packs and plans, no per-platform surcharge.
Start tracking AI Mode citations today
50 free credits on signup, no card. Your first request is a POST with a query in it.