Your AI Citation Has an 11-Day Shelf Life: What Last Week's AI Search Research Means for How You Get Found

Someone searches your name. Before they see your website, an AI summary tells them who you are.

Last week's research made that picture clearer. Below are the answers to the questions professionals are asking about AI search, drawn from five studies published between September 27 and October 3, 2026. Each answer links to its source.

How often does Google now show an AI Overview when someone searches a brand name?

Most of the time. Ahrefs analyzed 232.8 million search crawls and found that AI Overviews appeared in 82.91% of US desktop Google results for branded keywords as of September 29, 2026. The September average was 73.3%, up from 61.3% in July. In November 2025 the figure was 13.1%. (Ahrefs)

Does the AI Overview push my own website down the page?

Often, yes. In Ahrefs' sample, 76.2% of branded AI Overviews appeared in the first position, ahead of the brand's own website. (Ahrefs)

Where does Google's AI get its information about a business?

Largely from a short list of outside sources. In the same Ahrefs sample, Wikipedia was the most-cited external source (71.9% of citations), followed by YouTube (38.6%) and LinkedIn (22.8%). Your own profiles on those platforms are now part of how you are introduced. (Ahrefs)

How long does an AI citation last?

Not long. Profound tracked 883,000 pages across seven AI engines from September 2025 to September 2026. The median page lost half its peak citation share within 11 days, and the climb to that peak took about 42 days. (Profound)

Do all AI engines cite pages for the same length of time?

No. Profound found that Perplexity and Claude spread citations across weeks, while ChatGPT, Google's AI modes and Gemini concentrate them into days. A sustained page on Claude was 2.8 times as likely as one on ChatGPT to still hold half its peak share two months later. (Profound)

What makes a citation last longer?

Presence across several engines. Pages cited by five or more engines were 5.5 times as likely as single-engine pages to stay above half their peak after eight weeks. Profound sells AI visibility tracking, so read this as vendor research. The practical takeaway is that AI visibility needs a refresh schedule, not a one-time push. (Profound)

Do FAQ blocks, schema markup and fast page speed help me get cited by AI?

Not by themselves. A preprint by Discovered Labs co-founders Ben Moore and Liam Dunne analyzed about 2 million citations from ChatGPT, Claude, Google AI and Gemini and tested more than 60 page features. FAQ blocks, structured data and Core Web Vitals showed positive correlations that reversed or fell to zero once the analysis controlled for the domain the page was on. (arXiv preprint, AEO Wire summary)

What does predict whether AI cites a page?

Two things stood out. At the page level, the strongest predictor was how closely the page's wording overlaps with the language of real user prompts. At the site level, domain-level AI authority outweighed the next-strongest feature roughly sixfold. In plain terms: write the way people ask, and build a reputation the engines already recognize. The authors run an AEO agency and the sample was B2B SaaS, so the finding needs more testing outside that category. (arXiv preprint)

What is "agent experience" (AX)?

AX describes whether an AI agent can successfully fetch and read a business's own website while researching a buyer's question. A September 28 preprint separates it from answer engine optimization, which focuses on off-site citations. (arXiv preprint, AEO Wire summary)

Does an agent-readable website get recommended more often?

In that study, yes. Across 37,927 simulated buyer journeys and 1,056 real businesses, agent-ready businesses had answers built from their own pages 78% of the time versus 56% for the rest. They were clearly recommended 1.9 times more often and received answers that were 41% more accurate. Answers about sites that were not agent-ready were 3.7 times more likely to omit facts the buyer had asked for. This is a single preprint, and one co-author co-founded an AI-agent search company, so treat it as an early signal. (arXiv preprint)

What are Meta's Muse and Instinct, and do they show their sources?

They are personal-agent shopping tools. Muse passed 2.5 million downloads after its September 8 launch. Instinct is a rival agent built on Shopify's catalog. Profound's test of 50 prompts in one retail category found that neither names the websites it visits by default. (Profound)

How does a business get recommended by an AI shopping agent?

Per Profound: get products into Meta's and Shopify's catalogs, keep your site navigable with standard semantic HTML, and state clearly who each product is for. In Profound's test, Meta's catalog returned about 60 products from 30 merchants, while live browsing surfaced only 4 products from 2 merchant sites. The sample is small and retail-specific. (Profound)

Can I track clicks that come from AI Overviews?

Not reliably today. A marketer reported seeing a ?st_source=ai_overview parameter on links inside AI Overviews and AI Mode, which would allow site owners to separate those clicks. Google has not confirmed it, and Search Engine Roundtable could not reproduce it. (Search Engine Roundtable)

What should a solo practitioner or small firm do with all this?

Five moves follow from the research:

1. Search your own name in Google, ChatGPT, Gemini, Claude and Perplexity. Read what each says.

2. Correct the outside sources. Update your LinkedIn profile, and check Wikipedia and YouTube if you have a presence there.

3. Write in the language of real questions. Use the exact phrasing your prospects use, in plain words.

4. Keep your own site readable. Clear pages, standard HTML, and the facts a buyer would ask for stated directly.

5. Refresh on a schedule. With citations fading in days to weeks, a quarterly update is too slow for your most important pages.

Why does any of this matter beyond rankings?

Because people decide before they contact you. A prospect who is anxious about hiring a lawyer, a dentist or an advisor looks for reassurance first. In 2026, part of that reassurance comes from an AI answer. The distance between how uncertain they feel and how safe they feel choosing you is the Comfort Gap™, and AI summaries now sit inside it.

The goal is the same as it has always been: be known before you're needed, with a record that is accurate, consistent and easy for both people and machines to read.

Sources

• Ahrefs, AI Overviews on branded searches: https://ahrefs.com/blog/ai-overviews-on-branded-searches/

• Profound, citation half-life: https://www.tryprofound.com/blog/the-half-life-of-an-ai-citation-is-11-days

• Discovered Labs preprint (arXiv): https://arxiv.org/abs/2609.35077v1

• Agent experience preprint (arXiv): https://arxiv.org/abs/2609.34951v1

• Profound, Muse and Instinct: https://www.tryprofound.com/blog/the-era-of-personal-agents-muse-and-instinct

• Search Engine Roundtable, AI Overview tracking parameter: https://www.seroundtable.com/google-ai-overview-link-tracking-parameters-42219.html

• AEO Wire (aggregator used to find and verify items): https://anythingengineoptimization.com/archive/