---
title: "Google's AI features: official guidance, reports and the…"
description: "Why this lesson exists Google is still where most search journeys start in the UK, US, Gulf and Pakistan, and in 2026 it changed three things that every…"
url: https://optimizeall.com/learn/ai-search-optimization-geo/google-ai-features-guidance-and-controls
updated: 2026-10-05
---

AI Search Optimization: SEO for AI Overviews & Answer Engines · How AI search and answer engines work · lesson 3 of 17 · 16 min

# Google's AI features: official guidance, reports and the opt-out control

## Why this lesson exists

Google is still where most search journeys start in the UK, US, Gulf and Pakistan, and in 2026 it changed three things that every SEO needs to understand precisely: it published an official optimisation guide for its generative AI features, it added AI reporting to Search Console, and it gave site owners a switch to opt out of those features. It also clarified that its spam policies apply to attempts to manipulate AI answers. This lesson turns each change into decisions you can make.

## 1. The official guide (May 2026)

Google's *Optimizing your website for generative AI features on Google Search* guide says, in essence, that "AEO" and "GEO" are still SEO for Google:

- AI Overviews and AI Mode are **rooted in Google's core ranking and quality systems**. Pages must be indexed and eligible to be shown with a snippet to appear as supporting links.
- Focus on **unique, non-commodity content** that serves people, with clear organisation and good media.
- Things you **do not need** for Google's AI features: an llms.txt file (Google's crawlers treat it like any other text file), breaking content into small "chunks", AI-specific rewrites, or special schema.
- AI Mode uses **query fan-out**, issuing related searches across subtopics, so thorough coverage of a topic's facets gives more chances to be used.

Read the guide yourself before quoting it to clients; Google updates documentation without big announcements.

## 2. Search Console's generative AI performance report

Rolled out from June 2026 and available to all sites by early September 2026, the report shows **impressions** of your URLs inside AI Overviews, AI Mode and AI Overviews in Discover, broken down by **page, country, device and date**.

What it does **not** show: queries, clicks, CTR or position. Clicks from AI features remain inside your normal web search totals in the Performance report. So use the report to answer "which of our pages are being used in AI answers, where, and is that growing?" — not "how much traffic did AI send?".

A practical analysis:

| Question | How to answer |
|---|---|
| Which templates earn AI impressions? | Export pages; classify by URL pattern (guides, products, locations) |
| Is AI exposure shifting between markets? | Compare country breakdowns month on month |
| Are AI impressions cannibalising clicks? | For the same pages, compare the AI impressions trend with Performance clicks and CTR trends (correlation, not proof) |

## 3. The Search generative AI control (opt-out)

Following a UK Competition and Markets Authority requirement, Google introduced a property-level **Search generative AI control** in Search Console **Settings**. It began rolling out in June 2026 and Google said the global rollout was complete by the end of August 2026. When enabled for a property:

- The site's links and content are excluded from **AI Overviews, AI Mode and AI Overviews in Discover**.
- The site otherwise **continues to appear in regular Search results and Discover**.
- Google states the setting is **not used as a ranking signal** elsewhere in Search.
- It has **no effect on AI training** — that is governed separately by the `Google-Extended` robots.txt token (Module 4).

This is the first Google control that separates AI answer features from normal snippets. Before it, the only levers were `nosnippet`, `data-nosnippet`, `max-snippet` and `noindex`, which also cut normal search snippets.

## Should you opt out? A decision table

| Situation | Likely choice | Reasoning |
|---|---|---|
| Local service, SaaS, e-commerce, agency | Stay in | Being a supporting link or mention in AI answers is exposure to buyers; opting out removes you but not your competitors |
| Publisher whose answers substitute for visits | Test carefully | Compare AI impressions with click trends first; consider a trial on a separate property if your structure allows |
| Premium research brand | Mixed | Keep public summaries eligible; gate full reports |
| Regulated content where summaries risk misquoting | Consider | Accuracy risk may outweigh exposure; document the decision with legal/compliance |

Whatever you decide, **record it** (date, owner, reason, review date) in your AI access policy, because this setting is invisible to anyone reading robots.txt.

## 4. Spam policies now cover AI answers explicitly

In May 2026 Google updated its spam policies' definition of spam to include "attempting to manipulate generative AI responses in Google Search". In practice: buying or faking citations, planting hidden instructions for AI systems, mass-producing pages to capture every prompt variant, or manufacturing mentions are treated like any other spam. The durable strategy is the one this course teaches — be accessible, clear, corroborated and genuinely useful.

## Hands-on: a monthly Google AI review in 20 minutes

```text
1. Search Console > Generative AI report > last 28 days vs previous 28 days
   - Export pages; tag each by template and market.
2. Performance report > same pages > clicks, CTR (compare the same periods).
3. Settings > Search generative AI control > confirm status matches policy.
4. Manual spot-check: run 5 priority prompts in AI Mode in your main market;
   note which pages are supporting links.
5. Log one line per finding in the AI visibility tracker (Module 5).
```

If you manage many properties, check whether your Search Console API client exposes the new report before building automations — availability can lag behind the interface.

## Worked example: a UK recipe publisher weighs the switch

A UK food publisher (illustrative) sees AI impressions for its "how long to roast a chicken" style pages rise while clicks on the same pages fall. The editor wants to opt out. The SEO lead argues for evidence first: she tags pages into "quick-answer" and "deep-content" groups, finds clicks falling mainly on quick-answer pages, and notes that deep recipe collections still earn clicks from AI Mode supporting links. Because the control is property-wide, opting out would remove both groups. The team decides to stay in, invest in deeper content that answers can't fully substitute, and review quarterly — a documented, reversible decision rather than a panic switch.

## Common mistakes

- Reading the generative AI report as a traffic report — it has no clicks.
- Flipping the opt-out without recording it, then forgetting why AI impressions vanished.
- Blocking `Google-Extended` to "leave AI Overviews" — it doesn't.
- Assuming AI-specific tricks will survive the spam policies.

## Video lecture: Google's AI features: official guidance, reports and the opt-out control

Lecture coming soon · 13 chapters · about 9 minutes. Read the full transcript below.

1. Google's AI features in 2026
2. The guide in one line
3. Generative AI report
4. The opt-out control
5. Old levers vs new control
6. Example 1: Leeds accounting practice
7. Example 2: UK recipe publisher (illustrative)
8. Spam policies now cover AI answers
9. Common mistakes
10. Three monthly questions
11. Watch me do it: monthly Google AI review
12. The 20-minute monthly routine
13. Recap and try this now

## Lecture transcript

### Google's AI features in 2026

In twenty twenty-six, Google did three things that every SEO needs to understand precisely. It published an official guide to its AI features. It added AI reporting to Search Console. And it gave site owners a switch to opt out of AI Overviews and AI Mode. Get these wrong and you'll either misreport your results or remove yourself from answers by accident. In this lecture you'll learn what each change really does, and how to make a documented decision about the opt-out.

### The guide in one line

Start with the guide. Google's message is that for its AI features, AEO and GEO are still SEO. AI Overviews and AI Mode are rooted in core ranking and quality systems. To be a supporting link, a page must be indexed and eligible for a snippet. Google also lists what you don't need: no llms.txt, no chopping content into tiny chunks, no special AI rewrites, no special schema. Think of it like a restaurant guide that says, we judge the food. Redesigning the menu font for our inspectors won't help.

### Generative AI report

Next, the report. Search Console's generative AI performance report shows impressions for your pages in AI Overviews, AI Mode and AI Overviews in Discover, by page, country, device and date. Here's the key idea. It shows exposure, not traffic. There are no queries, no clicks, no click-through rate. Clicks from AI features still sit inside your normal web search totals. So use it to answer, which pages are being used in AI answers, and is that growing? Not, how much revenue did AI send us?

### The opt-out control

Now the control. Following a requirement from the UK's Competition and Markets Authority, Google added a Search generative AI control in Search Console settings. Turn it on, and your site's links and content are excluded from AI Overviews, AI Mode and AI Overviews in Discover, while you stay in regular results and the Discover feed. Google says it isn't used as a ranking signal elsewhere, and it doesn't affect AI training. Training is a different lever, the Google-Extended robots token. Two switches, two different jobs.

### Old levers vs new control

Before this control existed, your only levers were nosnippet, data-nosnippet, max-snippet and noindex. They limit how Google shows your content in AI features, but they also cut your normal search snippets, which can hurt clicks. So this is the first clean separation. That doesn't mean most businesses should use it. If you sell services, software or products, AI answers are where buyers compare options. Opting out removes you, not your competitors.

### Example 1: Leeds accounting practice

Worked example one, simple. A two-person accounting practice in Leeds hears about the opt-out on a podcast and asks if they should flip it. We look at the business model. They want enquiries from small business owners, who increasingly ask AI search which accountant handles contractor tax. Being cited is exposure they can't buy. There's no content they're trying to protect. Decision: stay in, record the decision, review in six months. Total time, fifteen minutes.

### Example 2: UK recipe publisher (illustrative)

Worked example two, harder, with illustrative details. A UK recipe publisher sees AI impressions rise on quick-answer pages, like how long to roast a chicken, while clicks on those same pages fall. The editor wants to opt out. The SEO lead tags pages into quick-answer and deep-content groups. Clicks are falling mainly on quick answers. Deep recipe collections still earn clicks from AI Mode supporting links. Because the control is property-wide, opting out would remove both. So they stay in, invest in deeper content that summaries can't replace, and review quarterly.

### Spam policies now cover AI answers

One more change you must know. In May twenty twenty-six Google updated its spam policies to say spam includes attempts to manipulate generative AI responses in Google Search. Buying or faking citations, hidden instructions aimed at AI systems, mass-produced pages for every prompt variant, manufactured mentions. All of it can be treated as spam. So here's the rule of thumb. If a tactic only works by deceiving a machine, it's a liability, not a strategy.

### Common mistakes

Common mistakes to avoid. Reading the AI report as a traffic report. Flipping the control without writing down who decided and why, and then forgetting when AI impressions vanish. Blocking Google-Extended in the hope of leaving AI Overviews. And trusting any AI hack that looks like it breaks the spam policies. Each one is easy to avoid if you keep a written AI access policy.

### Three monthly questions

Let's make the report practical, with three questions you can answer every month. First: which templates earn AI impressions? Export the pages and tag each one by URL pattern, like guides, products or locations. You'll often find that one type of page does most of the work. Second: is AI exposure shifting between markets? Compare the country breakdown month on month, which matters a lot if you sell in the UK, the Gulf and Pakistan at once. Third: are AI impressions eating into clicks? For the same pages, compare the AI impressions trend with clicks and click-through rate in the Performance report. That's correlation, not proof, but it tells you where to look closer.

### Watch me do it: monthly Google AI review

Watch me do it. I'll run the monthly Google AI review on a real property. First, I open Search Console, choose the generative AI report, and set the date range to the last twenty-eight days compared with the previous twenty-eight. I export the pages. Next, in a spreadsheet, I add a template column with a simple rule: if the URL contains slash guides, it's a guide; slash products, a product; slash locations, a location page. I pivot impressions by template and by country. Then I switch to the Performance report, filter to the same top twenty pages, and export clicks and click-through rate for the same periods. I put AI impressions and clicks side by side for each page group, and I write one sentence per group: exposure up, clicks flat, or exposure down, clicks down. Then I open Settings and check the Search generative AI control. It's off, which matches our written policy, so I note the date I checked. Finally, I run five priority prompts in AI Mode in our main market and screenshot the supporting links. The whole routine takes about twenty minutes, and the output is five lines in the AI visibility tracker, not a vague feeling.

### The 20-minute monthly routine

Here's a routine that takes about twenty minutes a month. Open the generative AI report and compare the last twenty-eight days with the previous twenty-eight. Export the pages and tag them by template and market. Check the same pages' clicks and click-through rate in the Performance report. Confirm the AI control's status in Settings still matches your written policy. Then run five priority prompts in AI Mode in your main market and note which pages appear as supporting links. Log one line per finding. If you manage many sites, check whether the Search Console API exposes this report yet before you build automation, because API coverage can lag behind the interface.

### Recap and try this now

Recap. The guide says AI features run on core Search systems, so fundamentals win. The report shows AI impressions by page, but not clicks. The control removes you from AI surfaces without touching regular results or training. Try this now. Open Search Console, check the generative AI report for the last twenty-eight days, export the pages, and tag them by template. Then check the control's status in Settings, and write one line in your AI access policy: our decision, the reason, and the review date.

## Key takeaways

- Google's 2026 guide says its AI features run on core ranking systems; llms.txt, chunking, AI rewrites and special schema aren't needed.
- The Search Console generative AI report shows impressions by page, country, device and date — no queries or clicks.
- The Search generative AI control removes a site from AI Overviews, AI Mode and AI in Discover without affecting regular results or training.
- Google's spam policies explicitly cover attempts to manipulate generative AI responses.
- Record any opt-out decision with owner, reason and review date.

## Try it

Open your Search Console generative AI report, export 28 days of pages, tag them by template, and write a one-line opt-out decision with owner, reason and review date.

- [Previous: The AI search landscape in 2026](https://optimizeall.com/learn/ai-search-optimization-geo/the-ai-search-landscape)
- [Next: Writing citable content: clear, quotable, verifiable](https://optimizeall.com/learn/ai-search-optimization-geo/writing-citable-content)
- [All lessons of AI Search Optimization: SEO for AI Overviews & Answer Engines](https://optimizeall.com/learn/ai-search-optimization-geo)