# How to earn AI search citations

## Practitioner's notes:

**Last updated:** October 2026

### Getting started with your AEO/GEO strategy

- [ ] Your SEO efforts can contribute to AI Overviews visibility, but they don’t guarantee citations. **Keep your SEO and GEO strategies separate** — especially since GEO is changing fast and many methods are still experimental. ([source](https://stacysbrain.com/cards/brightedge-ai-overview-citations-rank), [source](https://stacysbrain.com/cards/ahrefs-ai-citations-dont-rank))
- [ ] **Plan AI and search together** rather than as separate channels, since nearly nine in ten AI shopping journeys also pass through a search engine. ([source](https://stacysbrain.com/cards/similarweb-ai-shopping-journeys-also), [source](https://stacysbrain.com/cards/similarweb-ai-plus-search-doubles), [source](https://stacysbrain.com/cards/semrush-buyers-google-brands-see))
- [ ] AI splits searches by intent, so divide your AEO strategy into **two pipelines** — optimized separately **for facts and opinions**. For facts: turn your website into an AI-readable database. For opinions: target Reddit, Quora, and major review platforms. ([source](https://stacysbrain.com/cards/lily-ray-facts-go-brand-sites), [source](https://stacysbrain.com/cards/airops-ai-citations-come-community), [source](https://stacysbrain.com/cards/lily-ray-product-pages-chatgpt-fan))
- [ ] Pay attention to **top-of-funnel brand presence**: AI-powered search is used most at the very start of a purchase. ([source](https://stacysbrain.com/cards/mckinsey-use-ai-search-learn), [source](https://stacysbrain.com/cards/similarweb-ai-recommended-brands-win))

### Getting your brand into the model's recommendation shortlist

- [ ] Check that you are **not blocking AI crawlers** before anything else — blocking one removes the chance of being cited by that platform at all. ([source](https://stacysbrain.com/cards/loonis-blocking-ai-crawlers-kills))
- [ ] If you want more AI mentions, focus on building **topical authority** and **brand trust**. ([source](https://stacysbrain.com/cards/lily-ray-chatgpt-fan-out-queries), [source](https://stacysbrain.com/cards/lily-ray-chatgpt-domains-per-answer), [source](https://stacysbrain.com/cards/airops-ai-brand-mentions-come))
- [ ] ChatGPT decides which brands to consider before it even looks at web search results, by mapping connections between entities. Your brand name needs to sit as close as possible to your target category across many **high-authority sites**, so you’re included in the initial fan-out query. ([source](https://stacysbrain.com/cards/lily-ray-brands-chatgpts-fan-out), [source](https://stacysbrain.com/cards/lily-ray-chatgpt-injects-brands-unasked), [source](https://stacysbrain.com/cards/lily-ray-reddit-wikipedia-forbes-chatgpts))
- [ ] Keep your **key facts in plain HTML** on your own domain, so ChatGPT can retrieve them through site: searches. ([source](https://stacysbrain.com/cards/lily-ray-chatgpt-sometimes-guesses-wrong), [source](https://stacysbrain.com/cards/lily-ray-chatgpt-uses-site-fan))
- [ ] If your brand shares a name with other businesses or domains, consider adding **'Official Site'** to your homepage title or meta description so ChatGPT picks the right site. ([source](https://stacysbrain.com/cards/lily-ray-official-site-titles-helps), [source](https://stacysbrain.com/cards/lily-ray-chatgpt-sometimes-guesses-wrong))

### Making your pages quotable

- [ ] If you want a page to be cited by an assistant, consider **breaking the answer into a list or a table**. ([source](https://stacysbrain.com/cards/airops-lists-tables-chatgpt-cited))
- [ ] Optimize your pages for the classic **H1-to-H2-to-H3 heading hierarchy**. ([source](https://stacysbrain.com/cards/airops-clean-heading-hierarchy-lifts), [source](https://stacysbrain.com/cards/airops-structured-pages-earn-8x))
- [ ] Consider adding an **FAQ** block to your key pages. ([source](https://stacysbrain.com/cards/airops-pages-faqs-likely-cited), [source](https://stacysbrain.com/cards/pew-ai-summaries-long-google))
- [ ] **Refresh pages** that are already cited at least once a quarter, so they do not lose the position. ([source](https://stacysbrain.com/cards/airops-fresh-pages-under-months))
- [ ] Consider adding **JSON-LD schema** to pages that are not yet cited by AI. ([source](https://stacysbrain.com/cards/ahrefs-ai-cited-pages-3x), [source](https://stacysbrain.com/cards/airops-three-schema-types-lift))
- [ ] However, if you are adding schema purely to win more citations on pages that are *already* cited, it might not be the best tactic. ([source](https://stacysbrain.com/cards/ahrefs-adding-schema-ai-cited))
- [ ] If you want citations that **drive visits**, make sure the cited page leads somewhere: a product to buy, a quote to request, or research worth reading. ([source](https://stacysbrain.com/cards/semrush-ai-citations-drive-traffic))

### Getting your brand mentioned on sites AI quotes

- [ ] Most of what AI says about you is not on your site, so put part of the GEO effort into **mentions and links from third-party websites**. ([source](https://stacysbrain.com/cards/airops-ai-brand-mentions-come), [source](https://stacysbrain.com/cards/airops-third-party-ai-citations), [source](https://stacysbrain.com/cards/lily-ray-reddit-wikipedia-forbes-chatgpts))
- [ ] Focus on **best-of and comparison pages** on third-party sites that AI already cites — and on where you rank within them. This will likely outperform relying on your own comparison pages. ([source](https://stacysbrain.com/cards/airops-third-party-ai-citations), [source](https://stacysbrain.com/cards/lily-ray-ai-cites-brands-listicle))
- [ ] Consider reducing your focus on **self-promotional 'best of' listicles**, since Google’s AI Overviews are cutting their citations. ([source](https://stacysbrain.com/cards/lily-ray-ai-overviews-cite-fewer), [source](https://stacysbrain.com/cards/lily-ray-self-promotional-listicles-face), [source](https://stacysbrain.com/cards/lily-ray-ai-cites-brands-listicle))
- [ ] **Reddit’s share** of ChatGPT Search citations **collapsed** in August 2026 — so if the goal is ChatGPT visibility, it may no longer be the top priority for community work. ([source](https://stacysbrain.com/cards/promptwatch-reddits-chatgpt-citation-share))

### Optimizing for shopping AI agents

- [ ] Keep **ratings and review signals** prominent — this is the one signal every model relied on in every category tested. ([source](https://stacysbrain.com/cards/sabbah-acar-ratings-badge-helped-every), [source](https://stacysbrain.com/cards/sabbah-acar-social-proof-works-ai))
- [ ] Prioritize **improving the star score** over making urgency, discount or scarcity badges more extreme. ([source](https://stacysbrain.com/cards/sabbah-acar-badge-strength-barely-moved), [source](https://stacysbrain.com/cards/sabbah-acar-social-proof-works-ai))
- [ ] AI models tend to drop higher-priced products from recommendations, so consider **protecting price competitiveness** in AI-mediated commerce. ([source](https://stacysbrain.com/cards/sabbah-acar-higher-prices-made-ai))
- [ ] Since most models favor the **top-left product slot**, consider keeping that placement for priority products. ([source](https://stacysbrain.com/cards/sabbah-acar-llms-favored-top-left))
- [ ] Test **bundle badges** carefully with AI-agent traffic — in at least one model and category, they reduce a product’s chances of being picked by an AI assistant. ([source](https://stacysbrain.com/cards/sabbah-acar-bundling-backfired-gpt-phones))
- [ ] Write product copy that covers **multiple attributes** in one place, since shoppers ask assistants using complete requirements — not single keywords. ([source](https://stacysbrain.com/cards/similarweb-shoppers-ask-ai-several), [source](https://stacysbrain.com/cards/ahrefs-long-tail-query-variations))
- [ ] Improve your **purchase experience** for AI traffic so you don’t lose the lead: 60% of marketers agree the purchase experience should be as **frictionless** as the AI that brought them. ([source](https://stacysbrain.com/cards/hubspot-ai-visitors-need-frictionless), [source](https://stacysbrain.com/cards/semrush-ai-search-visitors-convert))

### Measuring the impact of AI optimization

- [ ] Expect buyers to reach you through **Google** or a **direct visit** after seeing you in an AI answer, so low AI referral traffic does not mean AI is not working. ([source](https://stacysbrain.com/cards/semrush-buyers-google-brands-see), [source](https://stacysbrain.com/cards/pew-click-link-inside-ai))
- [ ] Add AI **brand visibility** and **citations** to performance tracking. Until the full AI journey is traceable, use **branded search** and **direct traffic** as a proxy. ([source](https://stacysbrain.com/cards/iab-buyers-changing-measurement-ai), [source](https://stacysbrain.com/cards/pew-google-clicks-fall-when))
- [ ] In AEO/GEO, the scorecard that replaces SEO metrics includes: **citation rate**, **mention rate**, **share of voice**, **sentiment**, and **source attribution**. ([source](https://stacysbrain.com/cards/airops-five-metric-scorecard-proposed))
- [ ] Check visibility in **more than one model**. A brand that shows up in one can be missing from the next. ([source](https://stacysbrain.com/cards/airops-ai-brand-mentions-unique), [source](https://stacysbrain.com/cards/airops-brands-stay-visible-across))
- [ ] Track AI visibility **by language and use case** instead of relying on one blended mention rate. ([source](https://stacysbrain.com/cards/arxiv-brand-mention-rate-ai))
- [ ] Monitor your **site:** impressions in Search Console to see how often fan-out queries are reaching you. ([source](https://stacysbrain.com/cards/lily-ray-chatgpt-uses-site-fan))
- [ ] When estimating GEO impact, **separate the effect of the source change itself** from the downstream effect that flows through AI mentions. ([source](https://stacysbrain.com/cards/arxiv-source-state-term-geo))
- [ ] Consider tracking **AI referral traffic separately from organic** — it converts differently, and folding it into search hides the signal. ([source](https://stacysbrain.com/cards/hubspot-ai-referrals-have-buying), [source](https://stacysbrain.com/cards/semrush-ai-search-visitors-convert))

## Evidence layer

### How much of search is AI now: Understanding the market trends

- [Nearly half of consumers are already using AI-based search to guide their purchase decisions](https://stacysbrain.com/cards/mckinsey-nearly-half-consumers-use) — Tier D · McKinsey · Vendor report
- [Roughly a quarter of US, UK, and German consumers use AI tools to research shopping products](https://stacysbrain.com/cards/similarweb-quarter-us-uk-german) — Tier C · Similarweb (co-written with Statista) · Vendor report
- [74% of AI assistant users seek out AI-driven recommendations daily or weekly](https://stacysbrain.com/cards/kantar-ai-assistant-users-seek) — Tier D · Kantar · Vendor report
- [AI Overviews now appear for 48% of monitored queries, up from about 31% one year ago.](https://stacysbrain.com/cards/brightedge-ai-overviews-now-appear) — Tier C · BrightEdge · Vendor report
- [52% of queries tracked still have no AI Overview.](https://stacysbrain.com/cards/brightedge-tracked-queries-still-have) — Tier C · BrightEdge · Vendor report
- [Average AI Overview height rose by about 15% year over year, now exceeding a standard desktop screen and pushing organic results below the fold.](https://stacysbrain.com/cards/brightedge-ai-overviews-grew-taller) — Tier C · BrightEdge · Vendor report
- [11.4% of US desktop shopping sessions now involve an AI tool](https://stacysbrain.com/cards/similarweb-us-desktop-shopping-sessions) — Tier B · Similarweb (co-written with Statista) · Vendor report
- [Shoppers are more likely to visit the brand an AI recommended than a rival brand, even a week later](https://stacysbrain.com/cards/similarweb-ai-recommended-brands-win) — Tier B · Similarweb (co-written with Statista) · Vendor report
- [Generative-AI referral traffic to ecommerce sites jumped 203% year over year](https://stacysbrain.com/cards/similarweb-gen-ai-referrals-ecommerce) — Tier B · Similarweb (co-written with Statista) · Vendor report
- [AI alone produces the shortest shopping sessions; AI and search together produce twice as many site visits](https://stacysbrain.com/cards/similarweb-ai-plus-search-doubles) — Tier B · Similarweb (co-written with Statista) · Vendor report
- [Average AI search visitor converts 4.4 times better than organic search visitor](https://stacysbrain.com/cards/semrush-ai-search-visitors-convert) — Tier C · Semrush · Vendor report

### The converting power of AI search: What an AI answer does to the click

- [58% of marketers say AI referral traffic has more buying intent than organic search](https://stacysbrain.com/cards/hubspot-ai-referrals-have-buying) — Tier C · HubSpot · Vendor report
- [An AI summary appeared on 53% of Google searches of ten words or more and on 60% of searches starting with a question word, against 8% of one- or two-word searches](https://stacysbrain.com/cards/pew-ai-summaries-long-google) — Tier B · Pew Research Center · Independent research
- [People click a link inside the AI summary on just 1% of search visits with AI summaries](https://stacysbrain.com/cards/pew-click-link-inside-ai) — Tier B · Pew Research Center · Independent research
- [Google users ended their browsing session on 26% of pages with an AI summary, against 16% of pages with only traditional results](https://stacysbrain.com/cards/pew-sessions-end-ai-summary) — Tier B · Pew Research Center · Independent research
- [Google users clicked a traditional search result on 8% of visits where an AI summary appeared, against 15% of visits where none did](https://stacysbrain.com/cards/pew-google-clicks-fall-when) — Tier B · Pew Research Center · Independent research
- [49% of marketers agree that web traffic from search has dropped due to AI answers](https://stacysbrain.com/cards/hubspot-marketers-see-search-traffic) — Tier C · HubSpot · Vendor report
- [60% of marketers say visitors arriving from AI need a purchase experience as frictionless as the AI they came from](https://stacysbrain.com/cards/hubspot-ai-visitors-need-frictionless) — Tier C · HubSpot · Vendor report

### SEO vs. GEO: Rankings and citations seem to be different games

The two vendors in our database disagree sharply on **how much AI Overviews draw from the organic top 10**: one 2025 report puts it at 76%, while a 2026 report puts it at 17%.

We also have evidence pointing to **a decrease in AI Overview citations from the organic top 100**. That suggests this disagreement may reflect a real trend, and that the newer number is likely more reliable.

Still, both figures come from vendor reports, so treat the overlap as unsettled rather than known at this stage.

- [AI Overview citations pull from Google's top 10 pages 76% of the time](https://stacysbrain.com/cards/ahrefs-ai-overviews-cite-googles) — Tier C · Ahrefs · Vendor report
- [Optimizing for Google organic search remains the main way to get cited in major AI search products.](https://stacysbrain.com/cards/lily-ray-google-seo-still-main) — Tier D · Lily Ray · Practitioner
- [The share of AI Overview citations from sources not in the organic top 100 dropped sharply in some sectors, led by Entertainment where it fell from 92.2% to 46.6% in a year.](https://stacysbrain.com/cards/brightedge-ai-citations-outside-top) — Tier C · BrightEdge · Vendor report
- [AI assistants choose citations differently from Google or Bing, often picking pages from related queries rather than the exact same prompt](https://stacysbrain.com/cards/ahrefs-ai-assistants-cite-pages) — Tier D · Ahrefs · Vendor report
- [Only about 17% of AI Overview citations also rank in the organic top 10, while just over half are somewhere in the top 100.](https://stacysbrain.com/cards/brightedge-ai-overview-citations-rank) — Tier D · BrightEdge · Vendor report
- [AI Overview citations rarely match the organic top 10, with overlap steady around 17% overall, but some verticals like Healthcare see higher overlap at 24%.](https://stacysbrain.com/cards/brightedge-ai-overview-citations-overlap) — Tier C · BrightEdge · Vendor report
- [About 80% of AI assistant citations do not rank anywhere in Google for the original query](https://stacysbrain.com/cards/ahrefs-ai-citations-dont-rank) — Tier C · Ahrefs · Vendor report
- [When ChatGPT cites webpages, nearly 90% rank below 21st in Google](https://stacysbrain.com/cards/semrush-chatgpt-citations-rank-below) — Tier C · Semrush · Vendor report

### How ChatGPT's retrieval changed in 2026

- [ChatGPT decides which brands to consider before it even looks at web search results: Brands named in the initial fan-out query were cited in ChatGPT answers 33 times more often.](https://stacysbrain.com/cards/lily-ray-brands-chatgpts-fan-out) — Tier C · Lily Ray · Practitioner
- [ChatGPT now runs more fan-out searches per prompt, with free and cheaper tiers catching up to higher-tier models.](https://stacysbrain.com/cards/lily-ray-chatgpt-runs-fan-out) — Tier D · Lily Ray · Practitioner
- [ChatGPT now uses the site: operator in about 23% of fan-out searches, up from just 0.3% before.](https://stacysbrain.com/cards/lily-ray-chatgpt-uses-site-fan) — Tier C · Lily Ray · Practitioner
- [Product pages now make up 16.39% of retrieved pages in ChatGPT fan-out queries, overtaking listicles.](https://stacysbrain.com/cards/lily-ray-product-pages-chatgpt-fan) — Tier C · Lily Ray · Practitioner
- [The number of unique domains cited per ChatGPT response dropped from 19 to 15 after the 5.3 update, even as more pages are retrieved.](https://stacysbrain.com/cards/lily-ray-chatgpt-domains-per-answer) — Tier C · Lily Ray · Practitioner
- [The average number of fan-out queries per prompt jumped to 7.5 in ChatGPT 5.6 Sol, up from 2.17 in the 2025 baseline.](https://stacysbrain.com/cards/lily-ray-chatgpt-fan-out-queries) — Tier C · Lily Ray · Practitioner
- [The share of ChatGPT fan-out queries needing a second iteration rose from about 5% to 33.5% when 5.6 became default.](https://stacysbrain.com/cards/lily-ray-chatgpt-second-pass-fan) — Tier C · Lily Ray · Practitioner
- [For facts, ChatGPT fan-out queries go to official brand sites, while for opinions and product reviews, they favor Reddit and review platforms.](https://stacysbrain.com/cards/lily-ray-facts-go-brand-sites) — Tier D · Lily Ray · Practitioner
- [ChatGPT often injects brand names that the user never mentioned, across 11 of 13 product categories tested.](https://stacysbrain.com/cards/lily-ray-chatgpt-injects-brands-unasked) — Tier D · Lily Ray · Practitioner
- [ChatGPT sometimes guesses the wrong brand domain, directing site: queries at parked or unowned domains – especially for lesser-known brands.](https://stacysbrain.com/cards/lily-ray-chatgpt-sometimes-guesses-wrong) — Tier C · Lily Ray · Practitioner

### What makes a page quotable: Key facts

The two schema findings below come from the same publisher in the same year and point different ways: **schema correlates with citation, but adding it to pages** ***that are already cited*** **produced no lift.**

- [Structure boosts citation odds: clean heading hierarchies, FAQs, and schema earn citations at 2 to 2.8 times the rate of unstructured pages](https://stacysbrain.com/cards/airops-structured-pages-earn-8x) — Tier C · AirOps · Vendor report
- [Clean H1-to-H2-to-H3 heading hierarchy increases citation odds by 2.8x in AI answers](https://stacysbrain.com/cards/airops-clean-heading-hierarchy-lifts) — Tier C · AirOps · Vendor report
- [Pages with FAQs are 40% more likely to be cited in AI answers](https://stacysbrain.com/cards/airops-pages-faqs-likely-cited) — Tier C · AirOps · Vendor report
- [Pages with three or more schema types are 13% more likely to earn citations in AI answers](https://stacysbrain.com/cards/airops-three-schema-types-lift) — Tier C · AirOps · Vendor report
- [Lists and tables appear in nearly 80% of ChatGPT-cited pages, against 29% of Google's top organic results](https://stacysbrain.com/cards/airops-lists-tables-chatgpt-cited) — Tier C · AirOps · Vendor report
- [Optimizing for multiple long-tail query variations helps increase chances of being cited by AI assistants](https://stacysbrain.com/cards/ahrefs-long-tail-query-variations) — Tier D · Ahrefs · Vendor report
- [Pages updated within three months are three times more likely to be cited in AI answers](https://stacysbrain.com/cards/airops-fresh-pages-under-months) — Tier C · AirOps · Vendor report
- [Pages cited by AI are about three times more likely to have JSON-LD schema than those that aren't cited](https://stacysbrain.com/cards/ahrefs-ai-cited-pages-3x) — Tier B · Ahrefs · Vendor report
- [Adding JSON-LD schema to already AI-cited pages did not lift citations on Google AI Mode or ChatGPT, and their AI Overviews citations fell slightly against matched pages](https://stacysbrain.com/cards/ahrefs-adding-schema-ai-cited) — Tier A · Ahrefs · Vendor report
- [Blocking any of the main AI crawlers eliminates your chance to get cited by AI answer platforms, no matter the quality or schema of your content](https://stacysbrain.com/cards/loonis-blocking-ai-crawlers-kills) — Tier D · Loonis · Vendor report
- [AI citations only drive traffic when the cited page gives readers a clear next step](https://stacysbrain.com/cards/semrush-ai-citations-drive-traffic) — Tier C · Semrush · Vendor report

### Link building wins: Where the AI citations actually come from

- [85% of brand mentions in AI search come from third-party sources, not the brand’s own domain](https://stacysbrain.com/cards/airops-ai-brand-mentions-come) — Tier C · AirOps · Vendor report
- [48% of AI citations come from community Q&A sites, professional social, editorial platforms, and review platforms](https://stacysbrain.com/cards/airops-ai-citations-come-community) — Tier C · AirOps · Vendor report
- [Reddit, Wikipedia, Forbes, Merriam-Webster, Consumer Reports, Healthline, and Walmart are among the most cited domains in ChatGPT answers.](https://stacysbrain.com/cards/lily-ray-reddit-wikipedia-forbes-chatgpts) — Tier C · Lily Ray · Practitioner
- [For every batch of roughly 20 legal prompts tested, ChatGPT returned only .gov domain citations every time.](https://stacysbrain.com/cards/lily-ray-chatgpt-cited-gov-domains) — Tier C · Lily Ray · Practitioner
- [AI summaries in Google results cite Wikipedia and government sites more often, while standard search results cite YouTube more often](https://stacysbrain.com/cards/pew-google-ai-summaries-lean) — Tier B · Pew Research Center · Independent research
- [Nearly 90% of third-party AI citations come from listicles, comparisons and review sites, and 80% of cited brands sit in the first three positions on those pages](https://stacysbrain.com/cards/airops-third-party-ai-citations) — Tier C · AirOps · Vendor report
- [70% of the time, when a brand's own listicle is cited in a Google AI Overview, the brand is left out and the listicle is used to recommend a competitor instead](https://stacysbrain.com/cards/lily-ray-ai-cites-brands-listicle) — Tier D · Lily Ray · Practitioner
- [Self-promotional listicles can help short-term in AI search but are already being targeted for demotion by Google.](https://stacysbrain.com/cards/lily-ray-self-promotional-listicles-face) — Tier D · Lily Ray · Practitioner
- [Compared to June, Google’s AI Overviews now cite 38% fewer self-promotional 'best of' listicles](https://stacysbrain.com/cards/lily-ray-ai-overviews-cite-fewer) — Tier D · Lily Ray · Practitioner
- [Reddit's share of citations in ChatGPT Search dropped from a steady 3.83% to 0.52% between August 7 and August 17, an 86.4% fall in ten days](https://stacysbrain.com/cards/promptwatch-reddits-chatgpt-citation-share) — Tier B · Promptwatch · Vendor report
- [Reddit's citation share in Google AI Mode dropped by 30.5% from 2.22% to 1.54% over the reporting period, with a clearer dip at the end of July](https://stacysbrain.com/cards/promptwatch-reddits-ai-mode-citation) — Tier B · Promptwatch · Vendor report
- [Microsoft's security team documented companies hiding prompt-injection instructions inside "Summarize with AI" buttons, and classifies the technique as an attack rather than a growth tactic](https://stacysbrain.com/cards/lily-ray-microsoft-hidden-prompt-injection) — Tier D · Lily Ray · Practitioner

### Your brand’s AI visibility is unstable by default

- [68% of brand mentions in AI search are unique to a single model, so a brand visible in one assistant can be absent from another](https://stacysbrain.com/cards/airops-ai-brand-mentions-unique) — Tier C · AirOps · Vendor report
- [Only 30% of brands stay visible in an AI answer from one run to the next, and just 20% stay visible across five consecutive runs of the same prompt](https://stacysbrain.com/cards/airops-brands-stay-visible-across) — Tier C · AirOps · Vendor report
- [A brand's mention rate in AI answers differs by model, by language and by question type, so one overall visibility figure is not enough](https://stacysbrain.com/cards/arxiv-brand-mention-rate-ai) — Tier A · arXiv · Academic

### When an AI agent picks the product: Key optimization trends

- [89% of AI shopping journeys also involve a regular search engine](https://stacysbrain.com/cards/similarweb-ai-shopping-journeys-also) — Tier B · Similarweb (co-written with Statista) · Vendor report
- [Product ratings are the only badge that helped every AI model in every product category to get recommended more often](https://stacysbrain.com/cards/sabbah-acar-ratings-badge-helped-every) — Tier S · Jafar Sabbah and Oguz A. Acar · Academic
- [Social proof has a positive effect on AI visibility and worked in 14 of 16 AI model/category combinations during the experiment](https://stacysbrain.com/cards/sabbah-acar-social-proof-works-ai) — Tier S · Jafar Sabbah and Oguz A. Acar · Academic
- [Changing badge strength (urgency, discount, or scarcity) usually did little for AI recommendations, while the star score was the clearest exception — and the only change that actually boosted recommendations](https://stacysbrain.com/cards/sabbah-acar-badge-strength-barely-moved) — Tier S · Jafar Sabbah and Oguz A. Acar · Academic
- [Shoppers increasingly ask AI shopping assistants for several product attributes at once, not single keywords](https://stacysbrain.com/cards/similarweb-shoppers-ask-ai-several) — Tier B · Similarweb (co-written with Statista) · Vendor report
- [Higher prices consistently made every AI model less likely to choose a product for recommendation](https://stacysbrain.com/cards/sabbah-acar-higher-prices-made-ai) — Tier S · Jafar Sabbah and Oguz A. Acar · Academic
- [Most LLM models favored the top-left product slot, but GPT-5 did not](https://stacysbrain.com/cards/sabbah-acar-llms-favored-top-left) — Tier S · Jafar Sabbah and Oguz A. Acar · Academic
- [Bundling backfired for GPT-5 in the phone category](https://stacysbrain.com/cards/sabbah-acar-bundling-backfired-gpt-phones) — Tier S · Jafar Sabbah and Oguz A. Acar · Academic

### Measuring the effect of AEO/GEO efforts

- [Most marketers are making or planning changes to performance measurement for the AI web, mainly tracking brand visibility and citations in AI tools](https://stacysbrain.com/cards/iab-buyers-changing-measurement-ai) — Tier C · Interactive Advertising Bureau · Industry survey
- [For tracking AEO/GEO efforts, they propose a new scorecard built on five metrics: citation rate, mention rate, share of voice, sentiment and source attribution](https://stacysbrain.com/cards/airops-five-metric-scorecard-proposed) — Tier D · AirOps · Vendor report
- [Buyers often use Google Search or visit a cited website directly after seeing a brand mentioned in an AI answer](https://stacysbrain.com/cards/semrush-buyers-google-brands-see) — Tier C · Semrush · Vendor report
- [The source-state term made up 66.34% of the total GEO effect in the controlled design](https://stacysbrain.com/cards/arxiv-source-state-term-geo) — Tier C · arXiv · Academic

---

Source: https://stacysbrain.com/collections/how-to-earn-ai-search-citations