25 long-form guides to being found, cited and recommended by AI search—built from OpenAI, Google, Anthropic and Cloudflare documentation and the best available studies. Written to be cited, not skimmed.
Strategy 14 min
There is no form to fill in and no setting to switch on. There is a short list of things that demonstrably matter, a longer list of things people sell that don’t, and a way to tell them apart.
Read the research
Research 19 min
Most AI-crawler studies read robots.txt and stop there. We also requested each home page as OAI-SearchBot, PerplexityBot and Claude-SearchBot, then re-checked every refusal with two control requests to see which blocks are actually aimed at AI.
Read the research
Retrieval 13 min
Most “is my site visible to AI” checkers read robots.txt and stop. The blocks that actually hide sites sit in the firewall, the JavaScript and the headers. Here is how to test all five layers, and how far a test from your laptop can be trusted.
Read the researchStrategy 11 min
It usually isn’t your website. It’s the handful of lists, directories and review pages ChatGPT read before it answered, and those pages can be found, checked and worked on.
Read the researchStrategy 12 min
Most explanations of ChatGPT’s sources blend OpenAI’s own documentation with vendor guesses and forum folklore. We separated the three, graded every claim, and checked the robots.txt files of sites ChatGPT is said to rely on.
Read the research
Strategy 13 min
Most GEO advice repeats the same checklist. We graded 18 of its claims against what the platforms document and what the tests actually found.
Read the researchPlatforms 13 min
Most pages ranking for this were written before 15 September 2026 and describe a plan Cloudflare then changed. Here is what each setting does now, what happened to old settings, and how to test your own site.
Read the research
Retrieval 12 min
Three conditions make a page eligible for AI Overviews and AI Mode. One of them cannot be seen from outside your Search Console, and the setting most people blame, Google-Extended, isn’t one of them.
Read the researchPlatforms 9 min
In 2025 Cloudflare wrote this line into the robots.txt of more than 3.8 million domains, yet the big AI crawlers don’t document reading it. Here is what it says, what it doesn’t do, how it differs from the IETF’s Content-Usage draft, and whether yours needs one.
Read the research
Measurement 10 min
Every AI visibility number you can trust today comes from four free dashboards and your own enquiry form. Here is what each one measures, what none of them can, and a routine that holds up in front of a client.
Read the researchMeasurement 7 min
It is a label, not a tracker. If you saw it in a link, you can delete it. If you own the site, it is the most reliable AI-traffic signal you have, and it still undercounts.
Read the researchMeasurement 9 min
Google’s new channel is a good start and a poor total. Here is exactly what it counts, where the rest of your AI traffic hides, and a copy-paste channel group we tested on 13 October 2026.
Read the researchMeasurement 13 min
Google now reports how often AI Overviews and AI Mode show your links, and lets you leave them with one setting. Here is what the numbers mean, what they leave out, and what Exclude actually switches off.
Read the researchMeasurement 14 min
Every other free AI metric is a proxy. Bing’s report counts the times an AI answer actually cited your page, and shows the search phrases behind those citations. Here is how to set it up, what each number means, and how to turn a grounding query into a fix.
Read the research
Local 11 min
Your Google ranking doesn’t carry over. Each assistant reads a different local database, and only one of them belongs to Google.
Read the research
Ecommerce 10 min
Shopping assistants read feeds, not prose. Most of the confusion comes from rules that changed in March 2026, and documentation that hasn’t caught up.
Read the researchRetrieval 12 min
Meta and Amazon now run separate crawlers for search, training and live fetches. Two of them decide whether Meta AI and Alexa can cite you. Most robots.txt files we checked name the training bots and ignore the search ones.
Read the researchAgencies 11 min
Clients are asking for AI visibility faster than anyone can prove it works. The agencies that keep those clients sell two different things at two different prices, and report only what they can measure.
Read the researchRetrieval 8 min
OpenAI runs three agents with three different jobs. Treating them as one bot, in robots.txt or in a CDN setting, is how sites opt out of ChatGPT search without meaning to.
Read the researchSchema 7 min
Schema is not a citation lever. It is a machine-readable restatement of facts your page already shows, and in 2026 that is worth having for rich results and entity clarity, not for AI answers.
Read the researchSemantics 7 min
Google calls semantic HTML a good idea, not a requirement. For crawlers that don’t run JavaScript, the bigger question is whether your words are in the HTML at all.
Read the research
Performance 7 min
A green Lighthouse score won’t help if a firewall refuses the crawler or the content only exists after JavaScript runs. Speed matters most at the slow extreme.
Read the researchAuthority 7 min
Google’s systems reward pages that make it clear who wrote them and why they can be trusted. The AI-specific evidence is thinner than most guides admit, and the strongest of it points off your site.
Read the researchMetadata 6 min
Your canonical tells Google which URL to credit. Other AI engines don’t document their rule, so the safe move is one URL that every signal agrees on.
Read the researchOutput 6 min
A score nobody can act on is a vanity metric. Every SEO·AEO·GEO finding says what was checked, what was found, what to change and how strong the evidence is behind it.
Read the research