← Back to Blog
Updated March 29, 2026

Google Perspective API is Shutting Down: Your Complete Migration Guide (2026)

Last updated: March 2026

If you rely on Google's Perspective API for content moderation, you already know the clock is ticking. Google announced that Perspective API will be fully decommissioned on December 31, 2026. That gives you roughly nine months to find a replacement, migrate your integration, and verify everything works before the lights go off.

This guide covers everything you need to know: what is changing, what your options are, and how to migrate with minimal disruption.

What Was Perspective API?

Perspective API, developed by Jigsaw (a Google subsidiary), launched in 2017 as a free tool to detect toxic language in online conversations. It scored text on attributes like toxicity, severe toxicity, insult, profanity, threat, and identity attack, returning a probability score between 0 and 1.

It became widely adopted because it was:

  • Free with generous rate limits
  • Simple to integrate (single REST endpoint)
  • Reasonably accurate for English-language toxicity detection

Thousands of platforms, from small community forums to large news publishers, built their moderation pipelines around it.

Why Is Perspective API Shutting Down?

Google has not published a single definitive reason, but the shutdown aligns with several trends:

  1. Consolidation of AI efforts. Google is focusing its safety and moderation capabilities within Vertex AI and its broader Cloud AI platform rather than maintaining standalone free APIs.
  2. Maintenance burden. Perspective's models required ongoing retraining to keep up with evolving language patterns, and a free API does not generate revenue to fund that work.
  3. Better alternatives exist. The content moderation landscape has matured significantly since 2017. LLM-based approaches now offer richer, more nuanced classification than the original Perspective models.

Whatever the reasons, the result is the same: if you depend on Perspective API, you need a plan.

Shutdown Timeline

| Date | What Happens | |---|---| | Q1 2026 | Deprecation announced. Existing API keys continue to work. | | Q3 2026 | New API key issuance disabled. Rate limits on free tier reduced. | | December 31, 2026 | API fully decommissioned. All requests return errors. |

Do not wait until December. Migration takes time to test thoroughly, and you do not want to discover edge cases during the holiday traffic spike.

Your Migration Options

There are broadly three paths forward.

Option 1: Build Your Own

You can fine-tune an open-source model (such as a moderation-tuned variant of LLaMA or Mistral) and host it yourself.

Pros:

  • Full control over model behavior and thresholds
  • No per-request costs after infrastructure is set up
  • Data stays on your servers

Cons:

  • Significant engineering effort to build, deploy, and maintain
  • You own the model quality — retraining, bias testing, and drift monitoring are all on you
  • Infrastructure costs (GPU hosting) can be substantial
  • Time to production: weeks to months

This path makes sense if you have a dedicated ML team and content moderation is a core differentiator for your product.

Option 2: OpenAI Moderation API

OpenAI offers a moderation endpoint that classifies text across categories like hate, harassment, self-harm, sexual content, and violence.

Pros:

  • Free to use (included with any OpenAI API key)
  • Well-maintained by a well-funded team
  • Good accuracy for common categories

Cons:

  • Limited customization — you get OpenAI's categories, not your own
  • No direct mapping to Perspective's toxicity scores, so your existing thresholds break
  • Rate limits tied to your OpenAI plan
  • Requires an OpenAI API key and account

Option 3: Use a Drop-In Replacement Like moder8r

moder8r was built specifically for teams migrating off Perspective API. It provides a Perspective-compatible endpoint so you can switch by changing a single URL in your code.

Pros:

  • One-line migration (change the API base URL)
  • Returns the same toxicity attributes and score format as Perspective
  • Also supports expanded categories beyond what Perspective offered
  • Pay-per-request pricing with a free tier
  • No ML infrastructure to manage

Cons:

  • Third-party dependency (same as Perspective was)
  • Paid beyond the free tier

This is the fastest migration path if your goal is to keep your existing moderation logic intact while upgrading the underlying model quality.

Step-by-Step Migration to moder8r

Here is how to migrate from Perspective API to moder8r. The whole process can be done in under an hour.

Step 1: Create a moder8r Account

Sign up at moder8r.app. You will get an API key immediately. The free tier includes 10,000 requests per month, which is enough to test your integration thoroughly.

Step 2: Change the API Endpoint

If you are calling Perspective directly via REST, your current code looks something like this:

import requests

PERSPECTIVE_URL = "https://commentanalyzer.googleapis.com/v1alpha1/comments:analyze"

response = requests.post(PERSPECTIVE_URL, params={"key": GOOGLE_API_KEY}, json={
    "comment": {"text": user_comment},
    "requestedAttributes": {"TOXICITY": {}}
})

score = response.json()["attributeScores"]["TOXICITY"]["summaryScore"]["value"]

Change it to:

import requests

MODER8R_URL = "https://api.moder8r.app/v1/comments:analyze"

response = requests.post(MODER8R_URL, headers={"Authorization": f"Bearer {MODER8R_API_KEY}"}, json={
    "comment": {"text": user_comment},
    "requestedAttributes": {"TOXICITY": {}}
})

score = response.json()["attributeScores"]["TOXICITY"]["summaryScore"]["value"]

The response format is identical. Your downstream code that reads scores and applies thresholds does not need to change.

Step 3: Update Authentication

Perspective used a query parameter (?key=). moder8r uses a standard Authorization: Bearer header. This is the only structural change.

Step 4: Test with Real Data

Run a sample of your actual user content through the new endpoint and compare scores against your existing Perspective results. moder8r's models are more recent and may score some edge-case content differently. Adjust your thresholds if needed.

# Quick comparison script
for comment in sample_comments:
    perspective_score = get_perspective_score(comment)  # your existing function
    moder8r_score = get_moder8r_score(comment)          # new function
    delta = abs(perspective_score - moder8r_score)
    if delta > 0.1:
        print(f"Review: '{comment[:80]}...' | Perspective: {perspective_score:.2f} | moder8r: {moder8r_score:.2f}")

Step 5: Deploy and Monitor

Roll out to production. Monitor your moderation logs for the first week to catch anything unexpected. moder8r's dashboard shows request volume, average scores, and error rates so you can verify everything is healthy.

Feature Comparison

| Feature | Perspective API | moder8r | OpenAI Moderation | |---|---|---|---| | Toxicity scoring | Yes | Yes | No (binary categories) | | Score range | 0-1 | 0-1 | Boolean + score | | Perspective-compatible response format | -- | Yes | No | | Languages supported | 10+ | 15+ | 10+ | | Custom categories | No | Yes (Pro plan) | No | | Latency (median) | ~200ms | ~120ms | ~150ms | | Free tier | Yes (10 QPS) | Yes (10K req/mo) | Yes (with OpenAI key) | | Actively maintained | No (EOL 2026) | Yes | Yes |

Pricing

Perspective API was free. That is hard to beat on price alone. Here is what the alternatives cost:

| Provider | Free Tier | Paid Pricing | |---|---|---| | moder8r | 10,000 requests/month | From $0.001/request (volume discounts available) | | OpenAI Moderation | Free with API key | Free (but requires OpenAI account) | | Azure Content Safety | 5,000 requests/month | ~$1.50 per 1,000 requests | | Hive | Trial available | Custom pricing |

If your volume is under 10,000 requests per month, moder8r's free tier covers you at no cost. For higher volumes, the per-request pricing is competitive with other paid options and far cheaper than building and hosting your own solution.

Frequently Asked Questions

Can I keep using Perspective API until December 31?

Yes. Existing API keys will continue to work until the shutdown date. However, Google may reduce rate limits in Q3 2026, so do not assume your current throughput is guaranteed through the end of the year.

Will moder8r scores match Perspective exactly?

No. moder8r uses more recent models that are generally more accurate, but scores will differ on some inputs. Most teams find that score differences are small enough that their existing thresholds still work. We recommend running a comparison test (see Step 4 above) before going live.

What about Perspective's other attributes like IDENTITY_ATTACK and THREAT?

moder8r supports all six of Perspective's original attributes: TOXICITY, SEVERE_TOXICITY, INSULT, PROFANITY, THREAT, and IDENTITY_ATTACK. It also offers additional categories on paid plans.

Is there an official Google migration path?

Google has pointed users toward Vertex AI's content classification capabilities, but this is not a drop-in replacement. It requires rearchitecting your integration, uses a different pricing model, and has a steeper learning curve.

What if my volume is very high?

If you are processing more than 1 million requests per month, reach out to the moder8r team at moder8r.app/contact for volume pricing. Enterprise plans include dedicated support and SLA guarantees.

Start Your Migration Today

The worst time to migrate is the week before the shutdown. The best time is now.

  1. Sign up for a free moder8r account
  2. Swap your endpoint URL
  3. Run a comparison test
  4. Deploy with confidence

The free tier gives you 10,000 requests per month to test everything. No credit card required.

Have questions about migrating from Perspective API? Reach out at moder8r.app/contact.