Perplexity media results

Answers with media carry `videos` and `images` arrays next to the text. They matter most for how-to and product research prompts, where Perplexity often answers visually.

Returned in result

Request

curl -X POST "https://api.answerline.dev/v1/monitor/perplexity" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"Which project management tools do agencies use?","country":"US"}'

What Perplexity’s media results are

Media results are the videos and images Perplexity attaches to an answer. The API returns them as two arrays, videos and images, with the same item shape: a title, the media URL, a thumbnail, the source platform and dimensions.

A brand can be absent from the text and sources and still be the video a reader watches, or have its product photo shown for a comparison prompt. The media rate per prompt shows where video or image content can earn a slot, the share of items by source shows which platforms supply them, and each item’s url identifies the site behind it, so your own media can be tracked like a citation.

Field reference

result.videos[] and result.images[] share one item schema, MediaItem.

The medium enum includes map, so do not assume every item in images is a photo; filter on medium when the distinction matters.

Annotated excerpt

Values are illustrative; field names follow the POST /v1/monitor/perplexity result schema.

{
  "success": true,
  "result": {
    "text": "To replace a bike chain, first measure wear with a chain checker...",
    "sources": [
      { "position": 1, "url": "https://www.example-cycling.com/replace-chain", "label": "How to replace a chain", "description": "Step by step..." }
    ],
    "videos": [
      {
        "title": "Replace a bike chain in 5 minutes",
        "url": "https://video.example.com/watch/abc123",
        "thumbnail": "https://img.example.com/abc123/thumb.jpg",
        "medium": "video",
        "source": "video.example.com",   // platform string: normalize before counting
        "thumbnail_width": 320,
        "thumbnail_height": 180
      }
    ],
    "images": [
      {
        "title": "Chain checker tool on a drivetrain",
        "url": "https://www.example-cycling.com/images/chain-checker.jpg",
        "thumbnail": "https://img.example.com/t/chain-checker.jpg",
        "medium": "image",
        "source": "example-cycling.com",
        "image_width": 1200,
        "image_height": 800
      }
    ]
  }
}

Cost and include flag

There is no media include flag. Perplexity returns videos and images when it shows them, and the price is the same either way: 47.3 credits for an async task and 54.3 for a synchronous call. include.markdown, include.html and include.rawResponse add nothing. A media study pays for every prompt run, including runs that show no media, so budget on prompts times samples.

Example: platform share and media trigger rate

The script samples each prompt, then reports per prompt the fraction of runs that returned videos or images, and across all runs each platform’s share of videos and images and how often your own domain supplied them.

import os
import time
from collections import Counter, defaultdict
from datetime import date
from urllib.parse import urlparse

import requests

API = "https://api.answerline.dev"
HEADERS = {"Authorization": f"Bearer {os.environ['API_KEY']}"}
PROMPTS = [
    "how to replace a bike chain",
    "best commuter bike under $800",
    "what does a worn cassette look like",
]
SAMPLES = 5
DAY = date.today().isoformat()
MY_HOSTS = {"example-cycling.com"}


def run_tasks(tasks):
    resp = requests.post(f"{API}/v1/async/task/batch", headers=HEADERS, json=tasks, timeout=60)
    resp.raise_for_status()
    pending = {item["task"]["id"]: item["index"] for item in resp.json()["results"] if item["success"]}
    results = []
    while pending:
        time.sleep(10)
        for task_id in list(pending):
            body = requests.get(f"{API}/v1/async/task/{task_id}", headers=HEADERS, timeout=30).json()
            status = body["task"]["status"]
            if status in ("COMPLETED", "FAILED"):
                index = pending.pop(task_id)
                if status == "COMPLETED":
                    results.append((tasks[index]["payload"]["prompt"], body["response"]["result"]))
    return results


def platform(item):
    name = item.get("source") or urlparse(item.get("url", "")).hostname or "unknown"
    return name.lower().removeprefix("www.")


tasks = [
    {
        "taskType": "PERPLEXITY",
        "payload": {"prompt": prompt, "country": "US"},
        "idempotencyKey": f"bike-media-{p}-s{s}-{DAY}",
    }
    for p, prompt in enumerate(PROMPTS)
    for s in range(SAMPLES)
]
runs = run_tasks(tasks)

per_prompt = defaultdict(Counter)
share = {"videos": Counter(), "images": Counter()}
own = Counter()
for prompt, result in runs:
    per_prompt[prompt]["runs"] += 1
    for kind in ("videos", "images"):
        items = result.get(kind) or []
        per_prompt[prompt][kind] += bool(items)
        share[kind].update(platform(i) for i in items)
        own[kind] += sum(platform(i) in MY_HOSTS for i in items)

for prompt, c in per_prompt.items():
    print(f"{prompt:40} videos {c['videos'] / c['runs']:5.0%}   images {c['images'] / c['runs']:5.0%}")
for kind, counts in share.items():
    total = sum(counts.values()) or 1
    top = ", ".join(f"{name} {n / total:.0%}" for name, n in counts.most_common(5))
    print(f"{kind}: {top}   (yours {own[kind] / total:.0%})")

The helper keeps each task’s prompt with its result by mapping the returned task id to the batch index, so rates can be grouped per prompt. The source value is used when present; otherwise the host of url stands in for the platform.

Pitfalls

A prompt can show videos in one run and none in the next. Measure the trigger rate over several samples, with the market fixed, before concluding that a prompt “has” media; see answer volatility.

source is a free-text string. Lower-case it, strip www., and keep a mapping table for platforms that appear under more than one name.

Width and height fields may be missing on an item, so don’t filter on them without a default.

medium can be map. If your metric covers photos or videos only, count items whose medium matches.

thumbnail is a separate URL from the media itself and may be on a different host. Attribute content by url.

Related

• Perplexity engine guide: sync and async calls, webhooks, geo targeting and billing.
• Perplexity sources for citations and Perplexity shopping for product, place and hotel blocks.
• Monitor Perplexity with an API for the full response walkthrough.
• Prompt set design for choosing prompts that separate how-to, product and local intent.
• Brand protection and AI visibility tracking.

More Perplexity data

All Perplexity fields · API reference