# How to get recommended by ChatGPT

When a buyer asks ChatGPT "what's the best X for Y", the model builds an answer from sources it can access, then names a short list of brands. This guide explains how that shortlist forms and what you can change on your own site this week.

## First: understand what you are competing against

ChatGPT is not scanning the whole web in real time. Depending on the mode, it draws on:

- its trained knowledge (stable, slow to change),
- live search results it retrieves for the question,
- pages it can fetch and read.

So visibility is a function of three things: **what the model already knows about you**, **what authoritative pages say about you**, and **whether your site is readable**.

## The seven levers that actually work

### 1. Be readable by AI crawlers

If your robots.txt or firewall blocks GPTBot, OAI-SearchBot or ChatGPT-User, the model cannot fetch your pages. Check the block explicitly and allow it. This is the single most common self-inflicted invisibility.

### 2. Own a clear category sentence

Models group brands by category. If your site describes you as "a company committed to excellence", the model has nowhere to place you. Say plainly: "We manufacture industrial air compressors for export." One clear sentence beats a page of slogans.

### 3. Publish answer-shaped pages

Write pages that answer the exact questions buyers ask, as direct prose in the first two paragraphs. Models extract passages, so a self-contained answer is more quotable than an answer buried in a narrative.

### 4. Make facts verifiable

Certifications, years in business, capacity, service regions, guarantee terms. Facts are what a model can repeat without hedging. Vague claims are dropped or softened into "one of the options".

### 5. Be consistent everywhere

Your homepage, about page, product pages and directory profiles should agree on name, category and key facts. Conflicting facts make a model prefer a competitor it can describe confidently.

### 6. Earn third-party mentions

Models weight what independent sources say. Directory listings, industry publications, review sites and partner pages all feed the consensus. One authoritative mention often beats five pages of self-description.

### 7. Remove contradictions and dead ends

Broken product pages, inconsistent prices and conflicting contact details create doubt. A model that cannot resolve a contradiction will often omit you rather than risk being wrong.

## What does not work

- Keyword stuffing. Modern models ignore repetitive keyword text.
- Fake reviews or invented certifications. These poison trust and can be detected across sources.
- Publishing hundreds of thin AI-generated pages. Volume without specificity is noise.
- Hiding content behind scripts the crawler cannot execute.

## A 30-day plan

**Week 1 — access and baseline.** Unblock AI crawlers, publish `llms.txt`, and record a baseline: ask ChatGPT five buyer questions in your category and note whether you appear.

**Week 2 — pages.** Rewrite your two highest-intent pages into direct answer form. Add a factual "key facts" block.

**Week 3 — consistency.** Align your about page, product pages and directory profiles to one entity description.

**Week 4 — third-party.** Pitch one industry directory or publication for an accurate listing.

## How to measure progress

Ask the same five questions each week in a fresh conversation and record three numbers: how many answers name you, how many cite your domain, and whether the description is accurate. Trends matter more than any single run, because model answers are non-deterministic.

Bidvix automates this loop: it samples the same buyer questions across engines and reports mention rate, citation rate and competitor share of voice.

## Related reading

- [What is generative engine optimization?](/guides/what-is-generative-engine-optimization)
- [Why doesn't ChatGPT recommend my brand?](/answers/why-does-chatgpt-not-recommend-my-brand)
- [How do I know if AI cites my website?](/answers/how-do-i-know-if-ai-cites-my-website)