How to get your blog cited by AI assistants
AI assistants cite the pages that are easiest to quote: an answer in the first sentence, a single figure, a named subject, and facts a crawler can fetch. More buyers now ask an assistant before they search, so being that page puts your business in the answer.
What gets quoted
- Plain, confident statements with one figure are the easiest to quote. That is the reason for the voice in chapter 5.
- Answer first: a heading that asks the question, the answer in the first sentence, then the backing facts (prices, areas, timelines, guarantees).
- Write quotable passages. A quotable passage starts with the answer; names the subject instead of "we" or "this"; contains a specific fact (a number, place, date or price); covers one question per section; uses the customer's words in their language; and makes sense without the paragraph before it.
- Headings make sense with the body removed. Readers and AI engines both scan them.
- Comparisons, price lists, specifications and steps go in real tables and lists.
- Facts are accurate and clearly attributed. Outside data names its current, authoritative source in the text and links it, so an assistant can verify and cite it.
- Structure supports expertise; it doesn't replace it. Assistants prefer first-hand knowledge: what the job involves, what changes the price, what goes wrong.
- Teaching guides get cited; ads don't.
- Publish your prices and guides. When a business publishes no prices and no guide, an AI answer quotes a competitor's or a calculator site's prices instead.
- Publish a statistics page, so writers and AI search systems have one page of your figures to cite.

What doesn't work
- No shortcuts: no hidden text for AI, no keyword lists, no markup claiming what the page doesn't say. Google's AI features documentation says no special files, AI text files or markup are needed for AI Overviews or AI Mode.
- Ship a curated llms.txt and keep it current, listing the key pages and new posts; it costs minutes. Google doesn't use it and nothing confirms other assistants do, so it never replaces fixing pages. Paying for "AI schema" or llms.txt packages instead of fixing pages is a common mistake.
- Common mistakes: blocking every AI bot; key facts only in images, PDFs or script widgets; inconsistent names and phone numbers across directories; generic pages that never state prices, areas or timelines; tracking rankings only.
How assistants find you
- AI search is built on search. A page has to be indexed and eligible for a snippet. ChatGPT search, Perplexity and Copilot (grounded in Bing) use their own crawlers and indexes.
- When an assistant searches, it writes the answer from the live results. Without search, the model answers from its training, which is why consistent mentions across the web matter.
- Allow search and answer bots, and decide on training bots separately. In OpenAI's crawler documentation, OAI-SearchBot and ChatGPT-User fetch for ChatGPT answers and GPTBot gathers training data. In Google's crawler documentation, Google-Extended has no effect on inclusion or ranking in Google Search. Check that your CDN or firewall isn't blocking AI crawlers by default.
- Get mentioned, accurately, on the sources assistants already cite: directories, comparison pages, local and trade media, industry associations.
- AI crawlers don't run JavaScript: Vercel's crawler study found none of the major ones render it. Prices and facts sit in the initial HTML, machine-readable facts match the visible text exactly, and key pages can have plain Markdown versions for AI tools.
Keep your facts the same everywhere
Assistants cross-check what they find, and a business whose facts disagree across the web gets quoted less or quoted wrong.
- One name, address, phone, price list and service area on the site, the Google Business Profile, directories and social profiles, in every language.
- Update every copy of a fact on the same day it changes: an old price on one directory becomes the price an assistant repeats.
- Write the facts on the page in full sentences, not only in a table cell, so a passage lifted alone still names the subject.
- Date the facts that change, such as prices and timelines, so an assistant can tell the current figure from an old one.
Write the answer box
Every post opens with a short answer box under its first sentence, and it is the passage assistants lift most often.
- Two sentences, 55 words at most, that answer the post's main question completely on their own.
- One exact figure or one named thing in it, so the answer can't be mistaken for a general opinion.
- The same answer in every language, written natively, never shortened in one of them.

Measuring visibility
- Use a fixed set of 30 real buyer questions in each language your customers use. Run each one three times on ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews and AI Mode, recording the date, platform and language.
- Track: mention rate, citation rate, competitor share, accuracy of the facts quoted, and AI referral visits and enquiries in analytics (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com). One screenshot proves little; the trend does.
- Accuracy beats mentions. An answer with an old price sends the wrong customers. Trace a wrong fact to its source (an old directory, a stale profile, an unclear page) and fix it there.
- Search Console counts AI Overview and AI Mode clicks under the Web search type, with no separate line (Google's documentation), so combine it with question tracking and analytics.
- Make one person own "how assistants describe us" and run the test questions every month.
Where to start this week
Visibility in AI answers is built from the same steps as ordinary SEO, in a different order: access and accuracy first, then volume.
- Check crawler access with the list in chapter 8, and lift any block on search and answer crawlers.
- Compare your name, address, phone and prices on the site and in the five main directories of your trade, and fix every difference the same day.
- Add an answer box to the five posts with the most impressions.
- Write down your set of buyer questions and run it once, to get a baseline for the monthly comparison.
A month later, run the questions again and compare the mention and citation rates with the first run. You see which changes worked instead of guessing.
Edited by Andres LepikCo-founder & Head of Growth
Fact-checked by Siim SaarHead of SEO
Post by Rasmus Kivi
Co-founder & CEO
Co-founded REND and still reads every client's monthly report himself before it goes out.