When someone asks ChatGPT for "the best family dentist in Kochi" or "an invoicing app for freelancers under $20", they get a short list of names, not ten blue links. If your business isn't on that list, you don't exist for that customer. This guide explains how the list gets made and, more usefully, how to get on it.
We'll keep it honest. Nobody can guarantee placement in an AI answer, and OpenAI, Google and Perplexity all say as much. What you can do is make your business easy for an AI to find, verify and describe. That is what this guide covers.
Key takeaways
AI assistants don't rank pages. They answer questions by reading a handful of trusted sources and naming whoever those sources describe clearly. Your job has three parts: let AI crawlers in, state your facts plainly where they look, and get named on the third-party sites they already trust. Then measure whether it worked.
How AI assistants actually pick a business
Traditional search ranks web pages. AI assistants do something different: they run a search behind the scenes, read the top results, and synthesise an answer. The businesses that get named are the ones that appear consistently and unambiguously across those results.
ChatGPT's search feature, for example, combines the model with live web results and shows the sources it used inline, as OpenAI described when it introduced ChatGPT search. Perplexity is built around citations from the start. Google's AI Overviews and AI Mode are generated from Google's own index, and Google's documentation on AI features and your website is explicit that there's no special markup or opt-in beyond standard indexing.
That leads to a simple mental model:
- Retrieval. The assistant fetches a small set of pages for the question. Directories, review sites, comparison articles and your own site compete for those slots.
- Extraction. It pulls out entities (business names, products) and facts (location, price, services, ratings) from those pages.
- Synthesis. It names the businesses that are described most clearly and most consistently, and it often cites where it got each fact.
Every fix in this guide targets one of those three steps.
Step 1: Make sure AI can read you at all
This is the least glamorous step and the one most businesses fail. Each AI company runs its own crawler, and many sites block them without realising, either through a blanket robots.txt rule or a CDN bot filter.
| Assistant | Crawler to allow | What it's for | Documentation |
|---|---|---|---|
| ChatGPT search | OAI-SearchBot |
Surfacing your site in ChatGPT search results | OpenAI bots |
| ChatGPT (training) | GPTBot |
Model training; optional to allow | OpenAI bots |
| Perplexity | PerplexityBot |
Indexing for answers and citations | Perplexity bots |
| Claude | ClaudeBot |
Crawling for Anthropic's models | Anthropic help |
| Gemini / AI Overviews | Googlebot, with Google-Extended as the AI opt-out token |
Standard indexing; AI features use it | Google crawlers |
The important nuance: OpenAI separates search from training. You can block GPTBot (training) and still allow OAI-SearchBot (search visibility). Blocking both, which a lot of "block AI" robots.txt snippets do, removes you from ChatGPT search answers entirely.
Check your robots.txt right now. If you see User-agent: * followed by Disallow: /, or a specific block on any of the crawlers above, fix it first. Nothing else in this guide matters until AI can fetch your pages. We wrote a dedicated guide on which AI crawlers to allow and which to block.
Step 2: State the facts AI is looking for, in plain text
When an assistant reads your site, it's looking for a specific set of facts. If a fact isn't written in plain text on a crawlable page, the assistant either guesses, pulls it from a third party, or leaves you out.
For a local business, the facts that matter most:
- Exact services, each named the way customers ask for them ("Invisalign", not "clear aligner therapy")
- City and neighbourhood, in text, not just a map embed
- Opening hours, including weekends
- Price ranges or starting prices
- Who you serve (families, children, businesses)
- Reviews and rating, ideally with a count
For a SaaS or digital product:
- Who it's for, in one sentence
- Pricing, including whether there's a free plan and what it includes
- Integrations, each named individually
- Direct comparisons with the tools you get compared to
- Platform availability (web, iOS, Android, desktop)
- Limits and what's not included
A useful test: paste your homepage into a plain-text converter and read what's left. If your pricing lives in an image, your services are in a carousel that needs JavaScript, and your location only appears in a footer map, an AI reading the text version sees almost nothing.
Add structured data, but don't expect miracles
Schema.org markup helps machines confirm what your text says. For local businesses use LocalBusiness and its subtypes, following Google's local business structured data guide. For products use SoftwareApplication or Product with offers. Add FAQPage markup to genuine FAQ content.
Structured data is a confirmation signal, not a substitute for text. Our scans regularly find sites with perfect schema and no visible mention of the service in question. The schema alone doesn't get them named.
Step 3: Get named where AI already looks
This is the step that separates businesses that show up from businesses that don't, and it's the one most "AI SEO" advice skips.
Across our scans, AI assistants cite a small, repeatable set of third-party sources for each category. For local services in India that means Practo, Justdial and Google Business Profile. For SaaS it means G2, Capterra, Product Hunt, Reddit threads and "X vs Y" comparison articles. For agencies it's Clutch and case studies. We cover this in depth in where AI assistants get their answers.
The practical implication: if the top three sources for "best invoicing app for freelancers" are a G2 category page, a Reddit thread and a comparison article, and you're not on any of them, you won't be named no matter how good your own site is. The fix is to be present, accurately, on exactly those sources.
For local businesses, start with Google Business Profile. Make sure the category, services, hours and address match your website word for word. Inconsistency between sources is a common reason assistants describe a business wrongly or drop it.
Step 4: Write pages that answer the actual question
Assistants retrieve pages that match the question. A page titled "Services" doesn't match "dentist in Kochi with weekend appointments". A page titled "Weekend dental appointments in Kochi" does.
That means building specific pages for the questions customers ask:
- One page per core service, named the way people search
- A pricing page with real numbers, or at least ranges
- A "[Your product] vs [Competitor]" page for each major comparison
- An "alternatives to [Competitor]" page if you're the challenger
- A FAQ page that answers the questions in the customer's own words
These pages should be written to be quoted. Short paragraphs, one fact per sentence, the business name stated close to the fact. An assistant is far more likely to lift "Ledgerly offers a free plan for up to five clients and syncs with Stripe" than a paragraph of brand storytelling.
Step 5: Measure, wait, and measure again
Changes take time to show up. Technical fixes such as unblocking a crawler can appear in answers within days. New pages and listings usually take two to four weeks. Third-party mentions take longer still.
The only reliable way to know whether any of it worked is to ask the assistants the same questions, on a schedule, and record who they name. That's what TangentFlow does: it asks each assistant twenty of your customers' questions every week and tells you when one flips.
Why measuring matters more than ever
Clicks from search are falling as AI answers replace them. Pew Research found that Google users were less likely to click a link when an AI summary appeared, and an Ahrefs analysis estimated AI Overviews reduce clicks on top results by around a third. If the answer is the destination, being in the answer is the metric.
What doesn't work
A few things people try that our data doesn't support:
- Keyword-stuffing "AI" phrases. Assistants read for facts, not keyword density.
- Hidden text aimed at crawlers. It gets ignored at best and flagged at worst.
- Fake reviews or listings. These get removed, and the inconsistency they create hurts you.
- One-off "AI SEO" audits with no follow-up. Without re-checking, you have no idea what changed.
A 30-day plan
| Week | Focus | Actions |
|---|---|---|
| 1 | Access | Fix robots.txt, check CDN bot rules, verify the site renders as text, add schema |
| 2 | Facts | Rewrite service or product pages with explicit facts, add pricing, add a FAQ page |
| 3 | Presence | Claim and correct Google Business Profile, G2, Practo or whichever sources your scan cites most |
| 4 | Proof | Publish one comparison page, re-run your questions, note which flipped |
FAQ
Can I pay to be recommended by ChatGPT? No. There is no advertising product that places businesses in ChatGPT, Gemini or Perplexity answers as of this writing. Anyone selling guaranteed placement is selling something else.
Do I need a separate strategy for each assistant? Mostly no. They all reward clear facts and consistent third-party presence. The sources differ slightly: Perplexity leans on recent web pages, Gemini on Google's index and Google Business Profile, ChatGPT on Bing-style web results plus its own crawl. Measure each separately, fix the same things.
How many questions should I track? Twenty is enough to see patterns for a single business. Spread them across discovery ("best X in Y"), comparison ("X vs Z"), price ("affordable X") and specific-need questions ("X with weekend hours").
Does site speed matter? Only insofar as the crawler can fetch your page within its timeout. A slow but reachable site is fine. A site that times out for bots is invisible.
Sources
- OpenAI, Introducing ChatGPT search
- OpenAI, Overview of OpenAI crawlers
- Google Search Central, AI features and your website
- Google Search Central, Overview of Google crawlers and fetchers
- Google Search Central, Local business structured data
- Perplexity, Perplexity crawlers
- Anthropic, Does Anthropic crawl data from the web?
- Pew Research Center, Google users are less likely to click on links when an AI summary appears
- Ahrefs, AI Overviews reduce clicks
- Google, Guidelines for representing your business on Google
