Highlights
- The dataset: This report draws on over 220,000 AI source citations, collected with prompts specific to the marketing and advertising industry. The underlying AI responses, from four AI models, were gathered in 12 countries throughout Q2 2026.
- Two shared directories: Clutch.co and Sortlist.com are the only domains in all four models' top-20.
- Reddit makes up 8% of ChatGPT's citations: Its top source, ahead of every directory.
- Clutch.co alone makes up 6% of Grok's citations: Its single most-cited source.
- 70% of Google AI Overview's citations are from commercial domains: Versus 52% for Grok, the least commercial model.
- 51% local citation rate in Sweden vs 22% in Argentina: Local language matters far more in Northern Europe (39% overall across non-English markets).
- 20% average overlap between any two models: Roughly 80% of the top sources one model relies on are absent from another's top tier.
Which sources should you target to get cited as a marketing and advertising agency brand?
Each model's five most-cited sources, with their share of that model's total citations:
| Model | Top cited sources | Citation character |
|---|---|---|
| Grok | clutch.co (6.4%), sortlist.com (4.8%), agencies.semrush.com (4.6%), designrush.com (3.3%), digitalagencynetwork.com (1.5%) | Directories + listings |
| ChatGPT | reddit.com (7.5%), en.wikipedia.org (2.3%), awisee.com (2.0%), clutch.co (1.9%), theaustralian.com.au (1.7%) | Community + reference + directories |
| Microsoft Copilot | sortlist.com (4.7%), inbeat.co (2.0%), clutch.co (1.7%), linkatomic.com (1.6%), blissagency.it (1.6%) | Directories + agency sites |
| Google AI Overview | clutch.co (2.8%), agencies.semrush.com (2.0%), sortlist.com (1.7%), blissagency.it (1.6%), digitalagencynetwork.com (1.5%) | Directories + agency sites |
ChatGPT is the outlier at the top: Reddit is its most-cited domain by a wide margin and Wikipedia sits second, neither ranks high for the others. Grok is the most concentrated on directories, with its top four all being agency-listing platforms. All of these top sources appear in Temso's publisher directory.
How have AI source rankings changed over time in the marketing and advertising industry?
We split the observation window at its midpoint (early May 2026) and ranked the top domains by their share of citations within each half rather than raw counts.
Clutch and Sortlist held the top through both halves, while community and individual-agency sources (Reddit, blissagency.it) gained share and the global aggregators receded. Track the shifts in the live AI visibility rankings. With a single split, we read this as suggestive rather than conclusive.
What type of content do AI models cite for marketing & advertising agencies?
The models differ not just on which domains but on what type of content they trust. We classified every cited domain into six categories.
ChatGPT is again the most diverse: it cites the most reference content by far (7.7%, driven by Wikipedia) and the most editorial content (13.2%). Grok is the least commercial of the four, only 51.8% commercial, but a category-leading 29.2% UGC, reflecting how the large agency-listing and review directories it floods its answers with are classified, plus its Reddit and LinkedIn usage. Google AI Overview and Microsoft Copilot are the most straightforwardly commercial, sending roughly two-thirds of citations to agency and directory homepages.
Do AI models cite local-language content for marketing & advertising agencies?
Local sourcing varies sharply by country, and even the most local markets stay near half. Using the source's country-code top-level domain (ccTLD) as a proxy for local-language content, we measured the share of citations pointing to a local-ccTLD source in each non-English market.
A clear split emerges: the northern-European markets (Sweden, the Netherlands, Germany, Italy) all sit near half, while the three Spanish-language markets and France land in the low-to-mid 30s or below. Part of this is TLD habit, Spanish agencies frequently publish on .com, and part is that the big agency directories (clutch.co, sortlist.com, designrush.com) live on generic TLDs, so a Spanish or French prompt often lands on a .com directory rather than a national site. The ccTLD proxy therefore undercounts genuinely local content; treat these figures as a floor. The takeaway holds regardless: local language matters far more in Northern Europe than in the Spanish-speaking markets for this category.
Which AI model relies most on local sources for marketing & advertising agencies?
Holding the market constant, the models differ sharply in how local they go. Microsoft Copilot and Google AI Overview are the most aggressively local; Grok is the clear globalizer.
Google AI Overview reaches 80% in Dutch and 75% in German, and Microsoft Copilot 74% in German and 71% in Dutch, both reflecting a habit of surfacing national directories and local agency sites. Grok lands at or below 41% in every language and dips to just 18% in Spanish, because its citations are dominated by the big global directories rather than national ones. The country still drives localization more than the model does, but within a country Grok reaches for global aggregators while Google AI Overview and Microsoft Copilot reach for local ones. (Google AI Overview produced no in-scope French citations, shown as n/a.)
Should marketing & advertising agencies optimize for each AI model separately?
Yes, with two directories as the shared exception. When the four models answer the same agency questions, their top-cited domains barely converge. We compared the 20 most-cited domains for each model and measured how much each pair holds in common.
The reason the overlap isn't zero is structural: a small set of B2B agency directories anchors every model's list. Clutch.co and Sortlist.com appear in all four top-20s; DesignRush, agencies.semrush.com, and The Manifest recur across most. A marketing agency optimizing for one AI model is still absent from roughly 80% of the domains the other models prefer, getting rated on Clutch and Sortlist is the rare move that pays off across all four.
How many sources does each AI model cite per answer?
The models differ enormously in how many sources they cite per answer. Grok is by far the most source-hungry; ChatGPT the most economical.
For an agency, that cuts two ways: Grok offers far more citation slots to compete for, and many of them are directory listings, reachable by getting rated, but each individual citation carries proportionally less weight. ChatGPT cites few sources, so earning one of its six is both harder and more valuable.
Context
This analysis draws from Temso's AI visibility monitoring platform, which tracks how brands appear in AI model responses across ChatGPT, Microsoft Copilot, xAI's Grok, and Google AI Overview. The dataset covers marketing- and advertising-agency prompts across 12 markets (Argentina, Australia, Canada, France, Germany, Italy, Mexico, the Netherlands, Spain, Sweden, the United Kingdom, and the United States) and seven languages, between March 10 and June 29, 2026.
It measures which sources AI models cite, not which agencies or services are best. A high citation share means a domain is frequently surfaced by a model, not that it is authoritative or accurate. Citation behavior is a moving target: models update their retrieval and ranking continuously, so a snapshot like this captures a period, not a permanent state. Grok contributes the largest share of citations in this dataset.
Methodology
How we measured this
We analyzed every AI response and its cited sources for the marketing- and advertising-agency category over the observation window: over 220,000 cited source citations from 21,224 responses. A "citation" is a web source a model explicitly surfaced in support of an answer; we count only sources flagged as cited. Cross-model overlap was measured on each model pair's top-20 most-cited domains (shared domains divided by the combined set), averaged across all six pairs. Domain categories were assigned from a global domain registry. Temporal analysis split the window at its midpoint and ranked domains by within-half citation share. Source breadth is the mean cited sources per response, over responses citing at least one source, with a 95% confidence interval.
Localization is measured by a country-code-domain proxy: the share of citations on the local national ccTLD in each non-English market (.de for Germany; .ar/.mx/.es for the three Spanish markets). This proxy undercounts local content published on generic TLDs (.com), which are especially common among agency directories, so those figures are directional and conservative. The spread between the highest- and lowest-overlap model pairs was not statistically distinguishable at this scale (two-proportion z-test, p = 0.27), so we describe all four models as similarly low-overlap. Model coverage varied across the observation window, so temporal movements are reported as relative ranks within each half; the movement findings rest on a single midpoint split and are therefore suggestive rather than conclusive.
Frequently asked questions
Do different AI models cite the same sources for marketing-agency questions?
Mostly no, with two exceptions. Any two models share only about 20% of their top-20 most-cited domains on average, so roughly 80% of the sources one model relies on are absent from another's top tier. The exceptions are Clutch.co and Sortlist.com, the only two directories that appear in all four models' top-20.
Which single source matters most across all the AI models?
Clutch.co. It is the most-cited domain for Grok and Google AI Overview, sits in the top five of Grok, Google AI Overview, and Microsoft Copilot, and appears in all four top-20 lists. If there is one "must-win" source in this category, it is Clutch, followed closely by Sortlist.
Which sources does each model favor beyond the directories?
It varies sharply. ChatGPT leans on community and reference, Reddit is its #1 by a wide margin at 7.5% of citations, and Wikipedia is #2. Grok floods answers with directory and listing platforms at high volume. Microsoft Copilot and Google AI Overview send most of their citations to commercial agency and directory homepages.
Does local language matter for getting cited?
Yes, and it splits by region. In the northern-European markets, Sweden, the Netherlands, Germany, Italy, around half of citations go to local-ccTLD sources; in the Spanish-language markets (Argentina, Spain, Mexico) only about 22–30% do. Because the dominant directories live on .com domains, even local-language prompts often land on global sites.
Which AI model is most likely to cite a local agency's own website?
Google AI Overview and Microsoft Copilot. Google AI Overview reaches 75–80% local-ccTLD rates in German and Dutch, and Microsoft Copilot 71–74%; both surface individual agency sites more than the directory-heavy models. Grok is the opposite, it stays on global aggregators and rarely cites local sites, dipping to 18% in Spanish.
Which AI model cites the most sources per answer?
Grok, by a wide margin, about 29 cited sources per response, roughly 4.6 times ChatGPT's 6.3, with Google AI Overview at 9.5 and Microsoft Copilot at 8.1. More sources mean more slots to compete for in Grok, many of them directory listings you can earn by getting rated.
Did the sources AI models cite change over the period?
Modestly. Clutch and Sortlist held the top through both halves. Reddit and one Italian agency site (blissagency.it) gained share across models, while several global aggregators (GoodFirms, TechBehemoths, AgencyVista) slid down the ranking.
What should a marketing agency do with this?
Win the directories first, then optimize per model. Getting listed and well-rated on Clutch and Sortlist is the one move that pays off across all four AI systems. After that, treat each model as a separate channel: community presence (Reddit, Wikipedia) helps with ChatGPT, sheer directory coverage helps with Grok, and a strong, locally hosted agency website helps with Google AI Overview and Microsoft Copilot, especially in German- and Dutch-language markets.
