DEV Community: Anupam Pathak

We unified YouTube + Instagram data under one API key (no platform credentials needed) — here's the use case that drove us to build it

Anupam Pathak — Sat, 25 Jul 2026 04:47:01 +0000

We kept getting the same support request from developers:

"I'm building [influencer research tool / social audit SaaS / creator dashboard] and I need YouTube subscriber counts + Instagram engagement rates in the same response. What's the fastest way?"

The honest answer: there wasn't a clean one. You needed a Google Cloud project for YouTube, a Facebook developer account with app review approval for Instagram, separate OAuth flows for both, and you'd still hit YouTube's quota wall.

So we built Serpent's Social Media API to cover both under one key.

What's available:

YouTube:

/api/social/youtube/search — search with filters (type, duration, country, date, order)
/api/social/youtube/video — views, likes, comments, tags, description, thumbnails
/api/social/youtube/channel — subscribers, total views, video count
/api/social/youtube/playlist — list videos in a playlist

Instagram:

/api/social/instagram/profile — followers, following, engagement rate, recent posts, bio

No Google API key. No Facebook developer account. No OAuth. No quota limits.

Pricing comparison at 100K requests/month:

Platform Official API Serpent (Scale):

YouTube search Quota system (need approval) $1.00
YouTube video Quota system $1.00
Instagram profile Restricted (Graph API only for your accounts) $5.80

Quick code (YouTube + Instagram in same script):

import requests

BASE = "https://apiserpent.com/api/social"
HEADERS = {"X-API-Key": "YOUR_KEY"}

# Research a YouTube creator
channel = requests.get(f"{BASE}/youtube/channel", params={"id": "@mkbhd"}, headers=HEADERS).json()

# Research their Instagram
insta = requests.get(f"{BASE}/instagram/profile", params={"username": "mkbhd"}, headers=HEADERS).json()

print(f"YouTube: {channel['subscribers']:,} subscribers")
print(f"Instagram: {insta['followers']:,} followers")
print(f"Instagram Engagement: {insta['engagementRate']}%")

YouTube Data API alternative — no quota limits, no Google Cloud setup, from $0.01/1K

Anupam Pathak — Sat, 25 Jul 2026 04:28:24 +0000

If you've tried to build anything serious on top of YouTube's official Data API, you've probably been bitten by the quota system. Each search costs 100 units, you get 10,000 free/day, and requesting more requires Google's approval (which takes weeks and sometimes gets denied).

We added YouTube endpoints to Serpent API specifically because the quota wall kept killing legitimate tool ideas.

Endpoints:

GET /api/social/youtube/search — search videos, channels, playlists with filters (type, duration, country, order, safe)
GET /api/social/youtube/video — full metadata for any video (views, likes, comments, tags, description, thumbnails)
GET /api/social/youtube/channel — channel stats (subscribers, total views, video count)
GET /api/social/youtube/playlist — list videos in a playlist

What's different from the official API:

No Google Cloud project required
No daily quota limits
Flat, clean JSON (not Google's nested resource structure)
One API key for everything
Views and likes in the same response (Google makes you make separate calls)

Code:

import requests

# Search YouTube
resp = requests.get(
    "https://apiserpent.com/api/social/youtube/search",
    params={
        "q": "machine learning tutorial",
        "num": 10,
        "order": "viewCount",
        "duration": "long",
        "country": "us"
    },
    headers={"X-API-Key": "YOUR_KEY"}
)
results = resp.json()

# Get video details (views, likes, tags in ONE call)
video_resp = requests.get(
    "https://apiserpent.com/api/social/youtube/video",
    params={"id": "dQw4w9WgXcQ"},
    headers={"X-API-Key": "YOUR_KEY"}
)
video = video_resp.json()
print(f"{video['title']} — {video['views']:,} views, {video['likes']:,} likes")

I Compared What Google Shows vs What ChatGPT Says About My Brand. The Gaps Are Disturbing.

Anupam Pathak — Fri, 24 Jul 2026 12:39:26 +0000

Three months ago I ran an experiment that I haven't been able to stop thinking about.

I took 12 keywords relevant to my product — things like "cheapest SERP API," "Google search results API," "rank tracking API for developers" — and I checked two things:

Where does my product rank on Google for these terms?
What does ChatGPT say when someone asks about these topics?

The results were not the same. In some cases, they weren't even close.

For "cheapest SERP API," my site ranked on page one of Google. ChatGPT didn't mention me at all — it named three competitors, none of which are cheaper than us.

For "rank tracking API developers," we weren't on Google page one. But Perplexity cited our blog post in its top three sources.

Same product. Same content. Radically different visibility depending on which discovery channel your potential customer is using.

Why the divergence exists

Google's ranking algorithm optimises for signals it can measure: backlinks, authority, click-through rates, freshness, structured data. You can influence it directly.

LLM visibility is different. When ChatGPT answers "what's the best X?" it's drawing on training data weighted toward content that appeared frequently across the web before the training cutoff. Newer products with less historical footprint are structurally disadvantaged, even when they're genuinely better.

Perplexity, Claude with search, and Gemini with grounding add live web retrieval on top — which means they're a hybrid. They partly reflect training bias and current search results. Which means your SEO actually does affect your AI visibility, but not in a 1:1 way, and not uniformly across models.

This is why the divergence exists and why it matters: two customers, same query intent, using different discovery channels, will reach completely different conclusions about what's available in your market.

What I'm tracking now (and how)

I built a dual-tracking setup:

Layer 1: Traditional SERP tracking Standard rank tracking via SERP API. 12 target keywords. Daily checks. Costs roughly $0.22/month at my query volume.

# Daily rank check — traditional SERP
import requests

def check_serp_rank(keyword: str, target_domain: str) -> int | None:
    response = requests.get(
        "https://apiserpent.com/api/search",
        params={"q": keyword, "engine": "google", "num": 50},
        headers={"X-API-Key": "YOUR_KEY"}
    )
    results = response.json().get("organic_results", [])

    for result in results:
        if target_domain in result.get("url", ""):
            return result["position"]
    return None  # Not in top 50

Layer 2: AI visibility tracking Using the Serpent AI Rank API to query ChatGPT, Claude, Gemini, and Perplexity simultaneously for the same 12 keywords. Returns citation data — whether I'm mentioned, in what position, and what text is cited.

# AI citation check — same keywords, different channel
def check_ai_visibility(keyword: str) -> dict:
    response = requests.get(
        "https://apiserpent.com/ai-rank-api",
        params={"q": keyword, "engines": "chatgpt,claude,gemini,perplexity"},
        headers={"X-API-Key": "YOUR_KEY"}
    )
    return response.json()
    # Returns: visibility_score (0-100), citations, mentioned_text, position per engine

Two things jumped out immediately:

Claude and Perplexity cite me more than ChatGPT does. This probably reflects training data recency differences and how aggressively each model uses live retrieval vs. training memory.

My Google rank and my AI visibility are almost uncorrelated. A keyword where I'm #7 on Google might have me #1 on Perplexity. A keyword where I'm #3 on Google might have zero AI citations.

What actually improves AI visibility

I've been experimenting for 12 weeks. Here's what appears to move the needle:

What helps:

Getting cited in round-up posts on high-authority domains (Dev.to, Medium, Hashnode, TechCrunch). These get scraped into training data and also surface in live search retrieval.
Structured data on your pages (FAQ schema, product schema). Perplexity and Gemini especially seem to favour structured content.
Building content that answers direct questions, not just targeting head keywords. "What is the cheapest SERP API?" as a page title outperforms generic "SERP API" targeting for AI citation purposes.
Getting mentioned on GitHub READMEs and Stack Overflow answers. Both appear heavily in LLM training data.

What doesn't help much (yet):

Backlink quantity. High-DA backlinks improve Google. They appear to have weaker direct effect on AI citation than content structure.
Social signals. Twitter/X mentions don't seem to translate to AI citations in any measurable way.

Why this matters in 2026

AI-assisted search is not a future trend. It's current behaviour.

A developer searching for an API tool in July 2026 is as likely to ask Perplexity or ChatGPT as to type the query into Google. Possibly more likely.

If you're only tracking Google rankings, you have a blind spot that's growing every month.

The dual-tracking setup I described costs me less than $1/month in combined API calls. The insight it gives me is worth far more than that.

If you're building any kind of SaaS or developer tool and haven't checked your AI visibility yet: do it this week. You might find the same divergence I did.