Google AI Overviews SEO and AI Search Visibility Prompt Tracking: Official Guidance

Official guidance for Google AI Overviews SEO is Search Central, not a prompt-tracking vendor. AI search visibility prompt tracking is the extra log for the question GSC does not store: which wording produced the module, who else was cited, and what ChatGPT or Perplexity said on the same brief. We keep those jobs labeled. The reason we separate them is that the two jobs use different data, answer different questions, and report to different stakeholders. Mixing them produces a number that looks unified and is really two unrelated counts added together, which is the kind of KPI that fails the first time someone asks how it was built.

The documentation we follow is AI features and your website and Optimizing your website for generative AI features. The measurement layer for prompts is Promptwatch. A longer docs walkthrough is here. This post is the SEO-plus-prompt-tracking cut. The reason we point at the source docs first is that a vendor paraphrase drifts toward the vendor's product, while the Search Central pages stay still, and the still version is the one that holds up when a stakeholder asks where a claim came from.

What Search Central says you should do

Indexed. Snippet-eligible. Ordinary technical requirements. Google: no additional technical requirements for Overviews or AI Mode. The point of that sentence is that the baseline for appearing in a generative surface is the same baseline as appearing in classic Search, plus snippet eligibility. There is no separate Overview gate to engineer around, and a team that looks for one will spend time on a problem Google says does not exist.

Then the work we already sell: crawl allowed, internal links, page experience, important content in text, structured data that matches the page, Merchant Center and Business Profile where they apply. The optimization guide wants a unique point of view and non-commodity pages. Helpful content still applies. The reason this list looks familiar is that it is the same list a competent SEO program was already doing, and the AI surfaces reward the same foundation. A team that treated AI search as a brand new discipline often rediscovers that the basics were never finished in the first place.

What Google says to skip for Search: llms.txt, chunking into crumbs, a URL per fan-out variant, inauthentic mentions, special Overview schema. Schema is not required for Overviews. We still mark up what the user sees for rich results. The skip list is as important as the do list, because each item on it is something a vendor or a blog will recommend, and each one is a waste of engineering time. A URL per fan-out variant, in particular, is a content strategy that fragments authority across pages nobody would write on purpose, and it earns nothing from Google that a single well-written page would not earn.

June 2026 generative AI performance reports: impressions, pages, countries, devices, dates. Clicks are not in the launch list. We will not report impressions as clicks. The reason this is worth saying plainly is that the word "impressions" gets read as "traffic" in casual stakeholder conversation, and the two are not the same. An impression is a presence count. A click is a session. Reporting one as the other inflates the program and sets up a reconciliation that will not go well.

Where prompt tracking starts

GSC will not show the typed question, the competing citations in the module, or ChatGPT. Prompt tracking stores that row. The three things in that sentence are the three things a visibility program eventually needs and GSC cannot give. The typed question is the prompt that produced the module. The competing citations are the other URLs in the same answer. ChatGPT is the assistant surface that lives outside Google entirely. A prompt tracker exists to hold those three.

Paid Promptwatch logs AI Overviews and AI Mode next to ChatGPT, Gemini, Claude, and Perplexity, daily, from the UIs. Citation analytics name the URL. Essential at $95/mo covers that prompt and citation work, but it has no listed crawler-log allowance. Professional at $245/mo is the first brand plan with crawler logs, which show whether GoogleOther or the chat bots fetched the candidate. Kick-off at $199/mo also includes crawler logs. Business is $579/mo with 100 million logs. Explore is ChatGPT only and limited to 10 prompts, so it is not Overview prompt tracking. The way to read the plan ladder is that Essential answers the prompt and citation question, and the crawler log question opens at Professional for brands and Kick-off for agencies. A program that needs to know whether the bots fetched the page before it was cited has to be on a plan that includes logs, and Essential is not that plan.

Ahrefs Brand Radar is modeled and often Overview-adjacent. Semrush AI Toolkit is $99/mo per domain and modeled. Keep both. They are not official GSC, and they are not a frozen prompt list. Otterly $29 can include Overviews in its base set; Gemini/Claude add-ons lag 7 days. Profound Starter ($99 annual) is ChatGPT only. Peec is $95 for three models. Scrunch is $250 annual, weekly Excel. The reason we list these alongside Promptwatch is that a client often already pays for one of them, and the question is whether the engines it covers match the brief. Modeled tools are useful for a monthly view. They are not useful for a prompt-level citation row, and knowing which is which prevents a client from paying twice for the same approximation.

How we run Overviews SEO plus prompt tracking

1. Read both Search Central pages on the account. No ranking-factor folklore. The folklore is the thing that gets a program in trouble, because it sounds plausible and has no source, and a stakeholder who repeats it in a meeting will eventually be asked to defend it.
2. Confirm money URLs are indexed and snippet-eligible. Kill accidental nosnippet. Accidental nosnippet is a common silent failure, because the page is indexed and ranks, but Google is not allowed to show a snippet, which removes it from generative surfaces that depend on snippets.
3. Open GSC generative reports. Export impressions. Tab name includes "impressions." The tab name is a small thing that prevents a large error, because a tab labeled "impressions" will not be confused with a tab labeled "clicks" six months later.
4. Load the same buyer questions into Promptwatch so Overviews, ChatGPT, and Perplexity share a list. Sharing the list is what makes the two systems comparable, and a list that differs across tools produces two reports that cannot be read against each other.
5. Tickets: crawl first, passage second, third-party citation third. How to get cited by ChatGPT when the miss is an assistant. Google tickets stay Google tickets. The ordering is deliberate, because most visibility misses are crawl misses in disguise, and a team that jumps to the assistant layer without checking crawl usually finds the assistant was never given the page to cite.

Bing AI Performance is Copilot, not Overviews. Do not merge those exports. Ahrefs and Semrush still own classic ranks next to the module. The merge is the most common reporting error in this category, because both exports have the word "AI" in the column header and neither is the other one.

FAQ

Is prompt tracking required to do Overviews SEO?

No. Official SEO is eligibility plus helpful pages plus GSC impressions. Prompt tracking is required if the SOW also asks "which question" or "what did ChatGPT say."

Can we use Brand Radar instead of Promptwatch for Overviews?

As a monthly modeled view, yes. As official impressions, no (that is GSC). As a prompt-level citation row, no.

Should we add Overview schema?

Google says it is not required. We will not bill for a tactic the optimization guide already dismissed.

What to do this week

1. Bookmark the two Search Central URLs.
2. Open GSC generative reports. Export impressions.
3. Write 25 buyer prompts. Load Promptwatch Essential if you need the question-level Overview row plus ChatGPT.
4. Keep Ahrefs and Semrush running.
5. Email hello@1001seomedia.com if a stakeholder is treating modeled AI visibility as official guidance. We follow Search Central for Overviews SEO. We use Promptwatch when we need prompt tracking.