We respect your privacy.

We use strictly necessary cookies to keep you signed in and to protect against CSRF. With your permission we also use a small amount of first-party analytics to improve the product. We do not sell your data and we do not use third-party advertising trackers. See our cookie policy and privacy policy .

← All posts

Tag

#Geo

80 posts tagged Geo.

Google pays for grounding, not for links

Google's AI contribution pilot pays when a page shapes an answer, not when it is linked afterward. That rule says a citation count measures the wrong thing.

Sep 24, 2026 · 6 min · Crawlmind Engineering

A browser agent is not a crawler

Agentic browsing runs inside the user's own session, so robots.txt, bot allowlists and crawler analytics all miss it entirely.

Sep 23, 2026 · 5 min · Crawlmind Engineering

An MCP endpoint is not a discovery channel

NLWeb and MCP make your site answerable by agents that already found you. Nothing on the open web is hunting for a /mcp route yet.

Sep 22, 2026 · 4 min · Crawlmind Engineering

AI engines cite a different web

A 55,936-query study found 37% of the domains AI engines cite never show up in traditional results. The pool you compete in changed.

Sep 21, 2026 · 5 min · Crawlmind Engineering

Cloudflare flipped your crawler defaults

Cloudflare changed AI crawler defaults on September 15. Blocking training can also block Googlebot on any page that carries an ad.

Sep 21, 2026 · 5 min · Crawlmind Engineering

AI Mode is not AI Overviews with more text

Two independent studies put the URL overlap between Google's two AI surfaces near a tenth. One Google AI number hides two retrieval systems.

Sep 18, 2026 · 5 min · Crawlmind Engineering

Your ChatGPT ad does not buy the citation

ChatGPT ads sit under the answer, not in it. Advertisers are rarely cited in the answer above their own ad, so paid and earned visibility need separate plans.

Sep 17, 2026 · 5 min · Crawlmind Engineering

Claude finds you through Brave, not Google

Claude's cited sources track Brave's top 10 far more closely than Google's. Brave discovers pages differently, so a strong Google position may not carry over.

Sep 17, 2026 · 5 min · Crawlmind Engineering

IndexNow reaches Bing, not every AI engine

IndexNow pings reach a short list of engines. Google, OpenAI's crawler and Perplexity are not on it, so treat it as a Bing freshness tool, not an AI one.

Sep 17, 2026 · 5 min · Crawlmind Engineering

Front-load the page, not the paragraph

44.2% of ChatGPT citations come from the first 30% of a page, but 53% come from the middle of a paragraph. The two findings pull apart.

Sep 15, 2026 · 5 min · Crawlmind Engineering

Your local 3-pack does not carry into AI

Brands reach Google's local 3-pack 35.9% of the time and ChatGPT's recommendations 1.2% of the time. The gap is not ranking.

Sep 14, 2026 · 5 min · Crawlmind Engineering

Your PDF arrives with none of its signals

A PDF can be retrieved and cited, but it carries almost none of the metadata an HTML page uses to earn that citation.

Sep 14, 2026 · 6 min · Crawlmind Engineering

A citation is not a quotation

A working link to your page is the easy part. New evidence says roughly a third of AI citations distort the source they point at.

Sep 13, 2026 · 5 min · Crawlmind Engineering

Your sentiment chart is mostly noise

Brand mention barely moves between runs. The sentiment label flips about seven times more often, so most weekly tone charts measure nothing.

Sep 13, 2026 · 6 min · Crawlmind Engineering

Your audit stops at turn one

GEO tracking scores the first answer. A fifth of real AI search sessions keep going, and the follow-up turn re-decides who gets named.

Sep 13, 2026 · 5 min · Crawlmind Engineering

Agents quit where users would retry

A 10 second delay collapsed one web-agent benchmark from 42% to 2%. Agents get one pass at your site, with no second try.

Sep 12, 2026 · 6 min · Crawlmind Engineering

Your paywall hides the body, not the record

A paywall changes what an AI engine can read, not whether your URL can appear. The two failures land in different metrics, and the evidence disagrees.

Sep 12, 2026 · 5 min · Crawlmind Engineering

Your video is read, not watched

AI answers lean hard on YouTube, but the attributes that track with citation are description, timestamps and recency, not views or subscribers.

Sep 12, 2026 · 5 min · Crawlmind Engineering

Incumbency breaks at 0.075 stars

AI assistants default to the brand they already know, but only until one comparable number separates the options.

Sep 11, 2026 · 5 min · Crawlmind Engineering

Site moves break citations you can't redirect

A domain move redirects your pages. It does not redirect the listings, reviews and roundups that supply most of your AI citation volume.

Sep 11, 2026 · 5 min · Crawlmind Engineering

You are retrieved as a chunk, not a page

The retrieval unit is a passage the engine cut, not the page you wrote. Two 2026 studies show how much that boundary decides.

Sep 11, 2026 · 5 min · Crawlmind Engineering

AI visibility is measured on one persona

A persona audit cuts recommendation-set overlap by 0.12 to 0.20 Jaccard, and mid-market brands lose up to 75% of the set.

Sep 10, 2026 · 5 min · Crawlmind Engineering

Licensing deals don't buy AI visibility

OpenAI licensees earn 48% more citations per cited page on ChatGPT. The metric conditions on already being cited, and rival licensees show no premium.

Sep 10, 2026 · 5 min · Crawlmind Engineering

Your images reach AI answers as text

Alt text reaches only half of the major AI crawlers, and the multimodal GEO research shows caption text, not pixels, is what moves visibility.

Sep 10, 2026 · 5 min · Crawlmind Engineering

AI click loss now has a causal number

Two randomized field experiments measure what AI Overviews and AI Mode actually cost publishers. The two features cost very different things.

Sep 09, 2026 · 5 min · Crawlmind Engineering

Getting retrieved beats being retrieved first

A SIGIR 2026 reproduction fails to replicate lost-in-the-middle on modern models. What moves the answer is the company your page keeps.

Sep 09, 2026 · 6 min · Crawlmind Engineering

The fifth rerun buys almost nothing

A decomposition of 12,933 LLM brand answers shows language and model coverage cut measurement error far faster per query than extra reruns.

Sep 09, 2026 · 5 min · Crawlmind Engineering

Deep research cites you but gets you wrong

On deep research surfaces the link works and the page is relevant, yet the claim attached to it often is not supported. That gap is measurable.

Sep 08, 2026 · 6 min · Crawlmind Engineering

Your page is judged on a query nobody typed

Fan-out is two variables, not one number: whether search fires at all, and how wide it spreads. They move in opposite directions per engine.

Sep 08, 2026 · 5 min · Crawlmind Engineering

AI engines agree on brands, not on URLs

Engines agree on 6.8% of cited URLs but 30.3% of brand mentions. The layer you count decides whether you need one strategy or five.

Sep 07, 2026 · 4 min · Crawlmind Engineering

Reddit's citation collapse wasn't about Reddit

Reddit lost 86% of its ChatGPT citations in a week. The dates rule out the popular explanation, and that changes how you read your own drops.

Sep 07, 2026 · 5 min · Crawlmind Engineering

Most citation failures are relevance failures

A study of 949 cited-vs-uncited pairs puts only 10.1% of citation failures at the technical layer. The rest is relevance and presentation.

Sep 04, 2026 · 5 min · Crawlmind Engineering

Your own listicle recommends your rivals

Google AI Overviews cited self-promotional listicles but left the publishing brand out of the recommendation in 69% of cases.

Sep 04, 2026 · 5 min · Crawlmind Engineering

AI referrals land on your homepage

ChatGPT's brand-link update moved most AI referral clicks onto homepages. The page that earns the citation is no longer the page that gets the visit.

Sep 02, 2026 · 5 min · Crawlmind Engineering

AI visibility is per-language, not global

Query language picks the citation pool before your page is judged. What two multi-million-citation studies show, and why translation is no shortcut.

Sep 02, 2026 · 5 min · Crawlmind Engineering

Your AI bot policy now needs three answers

Cloudflare split AI traffic into Search, Agent, and Training, with new defaults landing September 15. One block rule no longer covers it.

Sep 01, 2026 · 5 min · Crawlmind Engineering

What Google and Bing's AI reports measure

Google Search Console and Bing Webmaster Tools now ship first-party AI visibility data. Two metrics, two denominators, two blind spots.

Sep 01, 2026 · 5 min · Crawlmind Engineering

Source diversity fails before accuracy does

A WWW '26 study shows synthetic content can take over 80% of top-10 retrieval while answer accuracy holds steady. Diversity collapses first, quietly.

Sep 01, 2026 · 5 min · Crawlmind Engineering

AI search has no appeals process

There is no ranking to inspect and no way to appeal an AI answer. What an ICML 2026 position paper says is missing, and what to do instead.

Aug 31, 2026 · 5 min · Crawlmind Engineering

Recognition is not recommendation

AI described 96% of tested brands accurately, yet 89% never showed up in category answers. Those are two different systems.

Aug 31, 2026 · 5 min · Crawlmind Engineering

What an AI licensing deal actually buys

OpenAI-licensed publishers see 48% more ChatGPT citations. The platform asymmetry inside the same study argues against reading that as bought placement.

Aug 31, 2026 · 5 min · Crawlmind Engineering

AI conversion lift: read the fine print

AI traffic went from converting 38% worse to 60% better in sixteen months. Most of that swing is measurement, not shopper behavior.

Aug 28, 2026 · 5 min · Crawlmind Engineering

Search Console's AI report is impressions only

Google's generative AI report shows impressions and nothing else. Read it as a coverage signal, not a performance metric.

Aug 28, 2026 · 5 min · Crawlmind Engineering

ChatGPT picks the site before the page

A fan-out change put domain-scoped searches ahead of page-level retrieval. Reddit lost 86% of its ChatGPT citations in four days.

Aug 27, 2026 · 5 min · Crawlmind Engineering

Your GEO rewrite can lose the retrieval

New end-to-end research finds the classic GEO tactics improve the answer stage while making pages harder to retrieve at all.

Aug 27, 2026 · 5 min · Crawlmind Engineering

AI citations are graded on a curve

Answer engines pick sources by comparing the candidates in context, not by scoring your page against a fixed bar. That changes what to fix.

Aug 26, 2026 · 5 min · Crawlmind Engineering

The crawl-to-refer ratio is not a verdict

It counts what a bot takes, not what it is worth. Why the metric cannot decide which AI crawlers you should block.

Aug 26, 2026 · 5 min · Crawlmind Engineering

Google's AI opt-out is not a crawler block

Google's new Search Console toggle decides whether you can appear in AI Overviews and AI Mode. It is a grounding switch, not a robots.txt rule.

Aug 26, 2026 · 6 min · Crawlmind Engineering

Cited but not absorbed: GEO's missing metric

Being listed as a source is not the same as shaping the answer. New research splits citation into selection and absorption.

Aug 25, 2026 · 4 min · Crawlmind Engineering

Formatting won't win you AI citations

A peer-reviewed 252,000-trial study found formatting-only edits barely move citation odds. Four gatekeeper factors decide instead.

Aug 25, 2026 · 5 min · Crawlmind Engineering

Half your AI citations aren't on your site

A study of 6.8 million AI citations found listings nearly matched first-party sites. Here is why, and where the finding does not apply.

Aug 25, 2026 · 5 min · Crawlmind Engineering

Why AI citation studies contradict each other

Three studies measured rank overlap in AI Overviews and got 38%, 54%, and 90%. The difference is the denominator, not the data.

Aug 25, 2026 · 5 min · Crawlmind Engineering

AI visibility is a distribution, not a number

Repeated runs of one prompt overlap by only about a third. Here is how many runs, prompts, and days a real AI visibility measurement needs.

Aug 24, 2026 · 5 min · Crawlmind Engineering

AI commerce: discovery outlives checkout

Checkout protocols churned hard through 2026. The product feed and the crawlable product page are the parts that kept working.

Aug 21, 2026 · 5 min · Crawlmind Engineering

Why your AI traffic still shows as direct

GA4 finally has an AI Assistant channel, but referrer mechanics still dump ChatGPT clicks into Direct. Why, and how to close the gap.

Aug 20, 2026 · 5 min · Crawlmind Engineering

Why original research earns AI citations

Data you publish yourself is uniquely quotable, and AI answer engines reward it. What the research shows, and how to package it.

Aug 19, 2026 · 5 min · Crawlmind Engineering

The click gap: what AI Overviews cost you

When an AI summary appears, organic clicks fall by half. Here is what the click gap means for how you plan and measure content.

Jul 06, 2026 · 5 min · Crawlmind Engineering

Query fan-out: answer the hidden queries

Google's AI Mode splits one question into many hidden sub-queries. Miss them and your page earns no citation. How to map and cover the fan-out.

Jul 06, 2026 · 4 min · Crawlmind Engineering

Why AI cites Reddit more than your site

User-generated forums outrank most brand pages in AI answers. Here is why, and the two-part strategy that actually works.

Jul 06, 2026 · 4 min · Crawlmind Engineering

From SEO reporting to GEO reporting

What breaks when you point an SEO dashboard at AI answers, and the metrics you add so the numbers mean something again.

Jul 05, 2026 · 5 min · Crawlmind Engineering

How to brief writers for GEO

A content brief that pushes writers toward citable answers instead of keyword density, so posts get quoted by AI engines.

Jul 03, 2026 · 5 min · Crawlmind Engineering

The cost of AI invisibility: a B2B framework

A qualitative way to reason about what your company loses when AI assistants answer buyer questions without ever mentioning you.

Jul 02, 2026 · 5 min · Crawlmind Engineering

Share of citation: the new share of voice

Share of citation measures how often AI answers name your brand. Here is why it is replacing share of voice as the metric that predicts growth.

Jul 02, 2026 · 5 min · Crawlmind Engineering

A repeatable AI-visibility content audit

A repeatable checklist for auditing whether AI assistants can find, read, and cite your pages, from crawler access to per-platform tracking.

Jul 01, 2026 · 5 min · Crawlmind Engineering

Block or allow AI crawlers: a decision guide

Whether to block or allow AI crawlers depends on one split: training bots versus search bots. A framework by business model.

Jun 30, 2026 · 5 min · Crawlmind Engineering

The atomic-answer pattern for AI citations

One question, one paragraph, one citable claim. The atomic-answer pattern is how you write content that AI engines can extract and quote cleanly.

Jun 29, 2026 · 5 min · Crawlmind Engineering

How answer engines deduplicate sources

Answer engines collapse many pages that say the same thing into one citation. Being the original, or adding new information, is how you survive the cut.

Jun 29, 2026 · 5 min · Crawlmind Engineering

When AI engines prefer tables over prose

AI engines reach for a table when the question is a comparison or a lookup. Here is when to use one and how to format it so the model can read it.

Jun 29, 2026 · 5 min · Crawlmind Engineering

Entity clarity: be legible to AI engines

Entity clarity is making your brand resolve to one consistent thing AI engines can recognize, name, and cite.

Jun 22, 2026 · 5 min · Crawlmind Engineering

Writing for follow-up questions in AI search

AI assistants now generate the next question for the user. Content that answers the follow-up, not just the headline query, gets cited twice.

Jun 22, 2026 · 5 min · Crawlmind Engineering

Why comparison pages win AI citations

Comparison and X vs Y pages match how buyers ask AI tools at decision time, which is why they earn citations. Here is how to build them.

Jun 20, 2026 · 5 min · Crawlmind Engineering

How freshness signals shape AI answers

Freshness signals tell AI engines how recently a page changed. Here is what dateModified and changelog pages actually do for citations.

Jun 19, 2026 · 4 min · Crawlmind Engineering

GEO vs AEO vs SEO: one funnel, three layers

GEO, AEO, and SEO are three optimization layers on one funnel. Here is how they differ, where they overlap, and how to stack them.

Jun 19, 2026 · 5 min · Crawlmind Engineering

Why answer-first structure wins AI citations

AI engines quote the part of your page that answers the question. Put that answer first and you get cited more often.

Jun 18, 2026 · 5 min · Crawlmind Engineering

Making a SaaS feature page quotable by AI

Product pages win AI citations at buying intent but lose at the research stage. Here is how to make a SaaS feature page liftable into an answer.

Jun 18, 2026 · 5 min · Crawlmind Engineering

How to measure GEO with citation share

Citation share is the percentage of AI answers on your topic that cite your site. Here is how to define it and track it.

Jun 18, 2026 · 5 min · Crawlmind Engineering

What structured data AI engines actually read

Structured data still helps AI engines extract your facts. But the schema landscape narrowed in 2025 and 2026, so the old advice is stale.

Jun 18, 2026 · 5 min · Crawlmind Engineering

The anatomy of an AI citation

What ChatGPT, Perplexity, and Gemini actually quote from a page, and the page traits that decide whether you are the source they lift.

Jun 17, 2026 · 5 min · Crawlmind Engineering

What is GEO, and how it differs from SEO

Generative Engine Optimization is structuring content so AI answer engines cite it. Where the term came from, what it means, and how it diverges from SEO.

Jun 17, 2026 · 4 min · Crawlmind Engineering

Document order beats heading level for AI

A page's heading outline is read in DOM order, not by level. Get the source order wrong and both crawlers and AI extractors misread your structure.

Jun 16, 2026 · 3 min · Crawlmind Engineering