Google AI Overviews vs ChatGPT: Half of ChatGPT's Holiday Citations Point to Pages Marketing Does Not Own

This edition compares two AI surfaces, Google AI Overviews and ChatGPT, on holiday shopping prompts. We pulled the top 50 cited rows per retailer on each engine, across four major US retailers. Both engines were filtered to the same holiday terms. Then we pulled the prompts themselves.

What this measures: This measures which pages each engine cites as sources for holiday questions. It does not measure what the answers say, and it does not measure clicks. Citation weight counts page citations, so a prompt that cites two newsroom posts counts twice. Every share in this edition is a share of citation weight. A share of distinct prompts is a separate pull and would give different numbers.

Shoppers bring different questions to each surface

Half of ChatGPT's top holiday prompts are logistics questions. On Google AI Overviews, 28% are logistics questions and 72% are product and deal queries.

Google AI Overviews gets keyword strings such as black friday tv deals and store hours thanksgiving. ChatGPT gets full sentences such as Is the store open on Thanksgiving? and What time does Black Friday start?

The average ChatGPT holiday prompt runs 6.0 words against 4.4 on Google AI Overviews. A keyword product query is answered by a category page. A logistics question is answered by a press release or a help article.

Each engine receives different questions and cites different pages, so each needs its own measurement.

Each engine cites a different mix of page types

Share of holiday citation weight by page type, Google AI Overviews against ChatGPT. Source: BrightEdge AI Hyper Cube. ChatGPT and Google AI Overviews, United States, June to August 2026. Four of six panel retailers, with both engines matched at the top 50 rows per domain, so both engines are compared at the same depth. Ratios use unrounded shares, so they differ slightly from ratios computed on the rounded values shown. Page owners are our characterization of a large retail organization. The shares are measured.

The last column names the engine with the higher share. ChatGPT gives more citation weight to newsroom posts, guides and help pages. Google AI Overviews gives more to commercial pages and on-site search results. Both engines cite all five types.

Commercial pages and on-site search results make up 74.4% of Google AI Overviews holiday citations and 35.5% of ChatGPT citations. Newsroom posts, guides and help pages make up 64.5% of ChatGPT citations and 25.6% of Google AI Overviews citations.

Two findings hold at every retailer in the pull. Commercial pages and on-site search results together run higher on Google AI Overviews for all four, by 10 to 53 percentage points. Help and policy pages run higher on ChatGPT for all four. The other three page types do not agree across all four. Those are panel averages, and the newsroom share on ChatGPT runs from 2% to 64%. Each retailer needs its own pull.

Pages owned outside marketing carry half of ChatGPT's holiday citations

In a large retail organization, marketing and merchandising own the commercial pages and guides. Those make up 72.1% of Google AI Overviews citations and 48.3% of ChatGPT citations. Communications and customer support own the newsroom and help pages, which make up 18.1% of Google AI Overviews citations and 51.5% of ChatGPT citations. On-site search results, owned by the ecommerce platform, make up the remaining 9.7% and 0.1%.

That changes who needs to be in the holiday planning meeting. On Google AI Overviews, updating the pages marketing owns covers most of the citations. On ChatGPT it covers about half.

Every retailer in the pull already has cited pages across at least four page types. Start by reviewing those cited pages for outdated information.

Example. Take an announcement post that carries last year's event date. If it is cited for the question of when a sale starts, anyone who opens it sees last year's date.

What this means for agentic commerce and AI shopping agents

These findings also matter for AI shopping agents. An agent needs four facts before it can buy: the sale date, the return window, the price-match terms and the shipping cutoff. Those four sit on newsroom posts and help pages, which carry 51.5% of ChatGPT's holiday citations.

The shopper typing a full sentence into ChatGPT today needs the same four facts an agent will need. Publishing them as pages a crawler can read gives both shoppers and AI systems more reliable information to work from.

On-site search results in Google AI Overviews call for a crawl and indexation check

Google AI Overviews cites retailers' own internal search result pages at 9.7% of its holiday citation weight. ChatGPT cites them at 0.1%. Two of the four retailers account for all of it, at 14.2% and 24.3%, and the other two show none. If a retailer's internal search pages appear in citations, the question for technical SEO is whether those pages should be crawled and indexed.

What to update before Black Friday

• Pull the cited-URL list for both engines. Group each list by directory and compare the two profiles. If they come back close to identical, check that the engine filter was applied before reading anything into the result.
• Share the ChatGPT list with the owners of those pages. The newsroom and help pages behind half of ChatGPT's citations sit with communications and customer support.
• Date every announcement and review last year's post. A cited announcement that carries last year's date shows that date to anyone who opens it. Sale start date, return window, price-match terms and shipping cutoff belong on pages a crawler can read.
• Update pages that already exist before commissioning new ones. Six weeks is a short window for new content to build visibility. Refreshing a page that is already live and already cited is the more practical priority.
• Plan around Google first and let the same page serve AI. Google still sends the traffic, so the commercial pages come first. They make up 64.6% of Google AI Overviews citations. Keep them current, then cover the newsroom, guide and help pages that make up 64.5% of ChatGPT citations.

How to check your holiday and agentic shopping visibility on both engines

Group the cited URLs by directory. In AI Hyper Cube, pull the cited-URL list for your holiday prompts and group it by directory. Each directory shows how much of the footprint comes from the newsroom, the help center and the guides.

Compare the two engines. Run the same pull on both and compare the two directory profiles. If they come back close to identical, confirm the engine filter was applied before acting on the result.

Size the organic footprint. Under the same question set, use Data Cube X to size the space before deciding how much to invest.

For the longer build program behind all of this, crawler access, product data and schema, see our guide to optimizing e-commerce for AI search.

Frequently asked questions

Does being cited mean I get the click?

This analysis does not measure clicks. It measures which pages are cited as sources.

Is this a share of prompts or a share of citations?

Citations. A prompt that cites two newsroom posts counts twice, so every share is a share of citation weight. A share of distinct prompts is a different pull and would give different numbers.

Should we stop investing in the Black Friday hub?

No. The hub sits inside the commercial pages that Google AI Overviews cites most, at 64.6%, and Google still sends the traffic. The finding is about which other page types carry the answer on ChatGPT.

How does this change what we do for agentic shopping?

It names the pages. An AI shopping agent needs the sale date, the return window, the price-match terms and the shipping cutoff. Those sit on newsroom and help pages, so those pages carry the agentic commerce work.

Why are the shares based on four retailers?

The page-type pull covers four of the six panel retailers, and the shares describe those four. They do not describe the other two.

Does this hold on other engines?

The pull covers ChatGPT and Google AI Overviews. Other engines are untested.

How current is this?

It reads June to August 2026. Holiday prompts run all year, so the pages cited in the summer are the pages already published going into November.

Technical methodology

Key takeaways

More from this series

See the rest of the Weekly AI & Search Insights series. For the build side of the same problem, read our guide to optimizing e-commerce for AI search.

BrightEdge tracks which pages ChatGPT and Google AI Overviews cite for your brand. Find out which holiday pages each engine cites for you before Black Friday with AI Hyper Cube, or request a demo.