How to get ChatGPT to recommend your brand (2026 playbook)

Someone in your category is about to ask ChatGPT which company they should hire. They will get three or four names and a paragraph on each, and they will contact one or two of them. This is the playbook for being on that list.

It is written for the four engines your buyers actually use — ChatGPT, Claude, Perplexity, and Gemini — because they behave differently enough that optimising for one alone leaves most of the opportunity on the table.

First, understand what you are optimising

When you ask ChatGPT “best branding agency in Bangalore”, it does not consult a ranking. It searches, reads a handful of pages, and writes a summary of what those pages say. The brands that get named are the ones that show up across the retrieved set with clear, consistent descriptions.

Which means the real question is not “how do I rank in ChatGPT”. It is:

For the exact question my buyer asks, which pages does the engine read — and am I on them?

You can approximate the answer in ninety seconds: type the buyer’s question into Google or Bing and look at the top ten results. That set is close to the retrieval set. Every page in it that does not mention you is a specific, fixable gap.

How the four engines differ

The practical read: Perplexity is the fastest to move — a new roundup listing can change its answer within a week or two. Gemini rewards your existing SEO, so if you rank on Google you are probably already there. ChatGPT is the stickiest, leaning on entity familiarity. And Claude is the most quality-sensitive — being in one substantial, credible source often beats being in five thin ones.

The playbook

Step 1 — Baseline before you touch anything

Write down five to eight questions a real buyer would type. Not your brand name — discovery questions:

• “Best [category] in [city]”
• “Who should I hire for [specific problem]?”
• “Top [category] compared”
• “Alternatives to [the market leader]”
• “Is [your brand] any good?” — the only one that names you

Run every one on all four engines and record who gets named, in what order, and which sources are cited. Keep the list — you will rerun it monthly, and a prompt set that changes between runs measures nothing.

Step 2 — Unblock the crawlers

Open yourdomain.com/robots.txt. If you see any of GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, or Google-Extended under a Disallow: /, that is the whole problem and nothing else on this list can work until it is fixed. Watch for the shared-group form, which is easy to miss:

User-agent: GPTBot
User-agent: ClaudeBot
Disallow: /

Many sites added these blocks during the 2024 scraping backlash and never revisited the decision. It was a defensible call then. It is an expensive one now.

Step 3 — Fix how your own site describes you

Three things, in order:

1. Server-render your content. Several AI crawlers do not execute JavaScript. If your positioning only exists after hydration, they see an empty page. Check by viewing source, not the inspector.
2. Add Organization JSON-LD with your legal name, city, founding year, and what you actually sell. This is how an engine confirms your identity instead of inferring it.
3. Make your homepage state the obvious. One sentence, near the top, in plain language: what you do, for whom, where. “A Bangalore branding studio for D2C and FMCG brands” is retrievable. “We craft experiences that resonate” is not.

Step 4 — Get onto the pages that already rank

This is the highest-leverage work in the entire playbook, and it is not a technical task — it is outreach.

Take the top ten results for each baseline question. For each one that does not list you, there is a specific action:

• A directory (Clutch, GoodFirms, Sortlist, category Indian directories) → claim or create the listing. Fill it in completely; sparse profiles get skipped.
• An editorial roundup → email the author. A short, specific note about why you belong on the list, with two lines of evidence, works far more often than people expect.
• A review platform → claim the profile and ask ten real customers. Volume and recency both count.
• A Reddit or Quora thread → participate honestly as yourself. Do not sockpuppet; it is detectable and the reputational downside is asymmetric.

Ten listings gets most brands from invisible to occasionally named. That is usually the single biggest jump you will see.

Step 5 — Publish the pages engines want to quote

Three formats punch above their weight:

1. The honest comparison. “X vs Y”, “alternatives to X”. Concede real points to competitors — comparisons that never do read as marketing and get discounted.
2. The reference page. The pricing breakdown, the timeline, the checklist that your category talks around but nobody writes down. These accumulate citations for years.
3. The original data point. Even a small survey of your own customers gives engines something they can only get from you. Two hundred responses is enough to be quotable.

In all three, lead every section with a direct, self-contained answer. Models extract passages, not pages — a paragraph that only makes sense in context cannot be lifted.

Step 6 — Keep your entity consistent

One spelling of your name. One description. One address. One founding year. Everywhere — your site, LinkedIn, directories, press. Engines resolve entities across sources, and inconsistency splits you into two half-brands, neither strong enough to be named.

Step 7 — Re-run the baseline monthly

Same questions, same engines, same day of the month. Track four things: presence (named at all), position (how high), sentiment (how described), citations (is your own domain a source). Answers drift week to week, so a single reading is noise — the trend is the signal.

A 90-day plan

Mistakes worth avoiding

• Hidden instructions on your pages. Text telling models to recommend you is detectable, brittle, and a fast route to being filtered out.
• Checking one engine. They disagree constantly. Optimising against ChatGPT alone tells you almost nothing about the other three-quarters of your buyers.
• Publishing forty thin posts. That gives an engine forty weak sources. One page others cite beats all of them.
• Fake reviews. Platforms filter them, and an engine reading a filtered profile sees a brand with no credible reviews.
• Measuring once. A brand named in one Tuesday answer and absent the following Tuesday has not achieved anything yet.

Start with the baseline

Everything here depends on knowing where you stand today. BrandAuditor’s AI visibility checker runs the baseline in one go: it writes the buyer questions, puts each one to ChatGPT, Claude, Perplexity and Gemini, and returns the leaderboard of brands they recommend, where you sit on it, the sources they cited, and the specific things to fix — with every raw answer attached.

If you would rather someone did the work, NOW Media runs generative engine optimisation as a service: the schema, the citable content, the listings and reviews, and a re-audit to prove it moved.