AI VISIBILITY FOR SMALL BUSINESSES  |  5 MODELS MEASUREDREPORT IN 3 DAYS
// sources

What we know about how AI picks who to name.

One post a week. Every claim carries a source you can open and check yourself — because a claim without a source is just an opinion with better clothes.

26 Oct 2026

What if AI gives customers your old opening hours?

Being named is little help if the answer sends someone to a closed door. Trace the wrong detail, correct the places that publish it and keep the website update separate from the next AI answer.

business information · accuracy · local businesses · source checking
25 Oct 2026

Are you asking AI the questions your customers actually ask?

A bike workshop can disappear from one AI answer and belong in the next. Before interpreting a visibility score, check what was asked, whose decision it represents and whether the question fits the business at all.

measurement · customer questions · local businesses · methodology
15 Oct 2026

Cited is not the same as named

A source link under an AI answer feels like a win. Semrush checked 3,981 appearances across four engines: in 61.7% the domain was linked and the brand never appeared in the answer text. ChatGPT links 87% of the time and names the brand in 20.7%; Gemini does the reverse. Only 11% of brands show up in all three models Fractl tested. How to check which half you are in.

citations · brand mentions · measurement · myths
1 Oct 2026

Does blocking GPTBot hide you from ChatGPT?

The advice is either "block the AI bots" or "allow them all", as if they were one thing. OpenAI documents separate crawlers for training and for ChatGPT search, and says each setting is independent. Only 0.08% of sites explicitly block the search bot, yet 5.54% turn it away through blanket rules; among top sites, 17.6% block GPTBot against 6.6% for OAI-SearchBot. What robots.txt can and cannot do, and the two-minute check.

crawlers · robots.txt · technical · myths
24 Sep 2026

Does adding statistics and quotes get you cited by AI?

The 2023 paper that named GEO found adding statistics, citing sources and quoting experts could lift AI visibility by 40% — the founding claim of the whole field. Two independent 2026 studies re-ran the test on today's engines: one found a Spearman correlation of 0.11 between content score and citation, the other found the same tactic cutting retrieval visibility by up to 36% in a full search-pipeline simulation. What gained visibility instead: titles, meta descriptions, schema — the boring stuff.

content strategy · myths · studies · structured data
18 Sep 2026

Nobody clicks the link

Every AI-visibility pitch promises traffic. Pew tracked 68,879 real Google searches: users clicked a link inside an AI summary on 1% of visits, and ordinary result links on 8% — against 15% with no summary. Cloudflare counted roughly 38,000 Anthropic crawls per referral sent back. The citation is still worth having, for a reason that shows up nowhere in your analytics: 88% of people who get an AI recommendation go and check it.

traffic · clicks · measurement · myths
10 Sep 2026

Can the AI even read your website?

On 858,457 small-business sites, 57% of AI crawler activity was a live fetch to answer a question, not a crawl. Two 2026 log studies show what that fetcher does: it reads the raw HTML only, ignores JavaScript and schema, and hangs up after five seconds (ChatGPT) or four (Gemini), then answers anyway. A third of crawler requests hit 404 pages. A two-minute check, and what to fix before anything about content.

crawlers · rendering · log data · technical
4 Sep 2026

Does a review profile get you cited by AI?

Seer Interactive sorted 1,926 brands by Trustpilot presence across 804,491 AI answers: no profile, 1% citation rate; as few as 1–13 reviews, 53.5%. Almost all of the gain sits between zero and a dozen. The study was published with Trustpilot and measures one platform — and SOCi's data says the star rating then works as a gate, not a dial. What that means for a firm without a marketing budget.

reviews · trust signals · local · myths
31 Aug 2026

Reddit was the answer. Then it wasn't.

Two datasets made Reddit the most-cited domain in AI answers, and the advice wrote itself. Then its share of ChatGPT citations fell from 3.83% to 0.52% in August, for a reason nobody has established. The much-quoted "vendors disagree on the baseline fourfold" turns out to be two different denominators, not a dispute — which is the more useful warning. Plus: Reddit is worth 24% of citations on one engine and 0.1% on another.

citations · platform risk · volatility · myths
27 Aug 2026

Do you need to rank on Google to get cited by AI?

Stuck on page three and told the AI will never mention you? Ahrefs put 15,000 questions to four assistants and compared the citations against the rankings: only 12% of cited URLs sat in Google's top 10, and about 80% did not rank for the question at all. Rankings still count — but the growth is happening at positions 21 to 100, and in some industries barely at all.

rankings · serp overlap · query fan-out · myths
20 Aug 2026

Does updating old content get you cited by AI?

Seer Interactive dated 7,683 pages behind 47,097 AI citations: 75% were updated within the last year. The twist is what counted as "fresh" — a quarter of those pages were first published over two years ago. Refreshing an old page beat writing a new one, and a second study found freshness the strongest citation predictor of sixteen measured.

content strategy · freshness · myths · updates
13 Aug 2026

Does link building get you cited by AI?

Every GEO pitch keeps the backlink line item from classic SEO. Two independent correlation studies say it barely applies: Ahrefs measured 75,000 brands and found Domain Rating — the backlink-based authority score — correlating just 0.266 with AI visibility, one of the weakest signals tested. YouTube mentions correlated at 0.737, the strongest single factor found. Where the outreach budget should go instead.

backlinks · correlation studies · myths · youtube
6 Aug 2026

Does AI-written content get cited?

AI now writes 52% of new articles online, so the pitch is to feed the machines content they wrote themselves. The data disagrees: 82% of the pages ChatGPT and Perplexity cite are still written by humans. Volume is the one move that reliably does not get you named — and why that is good news for a small team.

ai content · citations · content strategy · myths
30 Jul 2026

Does schema markup get you cited?

"Add structured data so the AI understands your page" is the most technical-sounding advice in GEO. Ahrefs added schema to 1,885 pages and measured against 4,000 controls: AI Mode +2.4%, ChatGPT +2.2% — both within noise. And five major AI systems ignore the code entirely when they read a page. Worth doing for Google, not the reason you'll get cited.

schema · structured data · myths · experiments
23 Jul 2026

The core and the carousel

Google's AI Mode swaps out 56% of its cited sources every week, ChatGPT 74% — and run the same query three times on one afternoon and only 9.2% of the URLs match. But underneath the churn sits a small core that never moves. Being named once tells you which half you were in, and it is probably not the one you want.

volatility · citation drift · measurement
18 Jul 2026

Does llms.txt do anything?

It is the first item on almost every AI-visibility checklist. Across 137,000 domains, 97% of llms.txt files received zero requests in a month — and the AI bots people publish it for accounted for 1.1% of the traffic that did arrive. What the log data says, and where the fifteen minutes should go instead.

llms.txt · crawlers · log data · myths
12 Jul 2026

Where AI answers actually come from

"Get listed in directories" is the advice everyone gives. The data says something more uncomfortable: ChatGPT pulled 58% of its local answers from business websites — and only 15% from directories. Here is what that means for a company with five employees.

citations · local · third-party mentions
// next post: Thursday, 29 Oct