The 5 Steps to Get Your Brand Recommended by ChatGPT

Step 1: Audit your current AI presence

Before writing anything, find out where you already stand. Run your buyers' actual questions through ChatGPT, Perplexity, Gemini and Google AI Overviews and record who gets cited. Most legal and compliance firms doing this for the first time are surprised to find a directory listing or a competitor's blog post answering questions they assumed only they could answer credibly.

This discovery step alone often reshapes the whole plan — a firm expecting to compete against three named rivals sometimes finds a generic legal-directory listing outranking all of them in AI answers, which changes the content priority from “beat the competitor” to “out-specify the directory.”

A proper audit also checks whether your site is even crawlable by the bots that feed these answers — if it isn't, every later step is wasted effort.

For a legal or compliance firm specifically, the audit should also flag any outdated jurisdictional claims already circulating in AI answers — a firm that expanded its practice areas or exited a market needs to know if AI models are still citing the old scope, since correcting a wrong existing citation is a different task than earning a new one.

Step 2: Fix crawler access first

Check robots.txt for OAI-SearchBot, ChatGPT-User, PerplexityBot, Google-Extended and Claude-SearchBot. These are the citation crawlers; blocking them makes your content invisible to AI answers no matter how well it's written. This is separate from the decision to block training crawlers like GPTBot or ClaudeBot, which is a legitimate policy choice about your content being used to train models rather than cited in real-time answers — the two decisions shouldn't be conflated.

This distinction trips up a surprising number of technically sophisticated teams. A developer following generic advice to "block AI bots" for training-data concerns will sometimes disable citation crawlers along with training crawlers in the same sweep, unintentionally opting the brand out of AI visibility entirely rather than making a deliberate, informed choice about training data alone.

Step 3: Restructure content around real prompts

Take each priority prompt and build a page that answers it directly in the first two sentences, then supports the answer with a comparison table and named specifics. Resist the temptation to answer several related prompts on one long page — a page trying to serve five prompts usually serves none of them clearly enough for a model to extract a confident, specific answer. A legal firm targeting "which firm handles cross-border data compliance for fintechs" needs a page that says exactly that, with named jurisdictions and service scope — not a general "about our practice" page that never states the answer plainly.

This same discipline extends to secondary but still valuable prompts, like naming the specific certifications or bar admissions relevant to the jurisdictions served. A page that states these plainly gives an AI model a concrete, quotable fact rather than the vague authority language that firms often default to in their marketing copy.

• One page, one primary prompt
• Direct answer in the opening two sentences
• A comparison or checklist table supporting the claim
• Named specifics (jurisdictions, certifications, timelines) instead of adjectives

Step 4: Test weekly and iterate

Re-run your fixed prompt set every week. Track which prompts you win, which a competitor wins, and what changed on their page that might explain it. Keep a simple log rather than relying on memory — three months in, it is easy to forget which prompt you were winning in January and lost by March without a written record to check against. This step is the one most in-house teams skip after month one, which is exactly when citations start drifting back to competitors who kept testing.

A simple way to keep this step from lapsing is to assign it a fixed calendar slot, treated with the same seriousness as a recurring client meeting, rather than leaving it as a task someone does when there's time. Teams that formalize the cadence sustain it; teams that don't, stop by month two almost without exception.

Step 5: Layer paid placement where it's live

A strategy also needs a stated timeline for when paid spend enters the plan, rather than an open-ended “eventually.” Once organic groundwork is in place, ChatGPT Ads (self-serve in the US since May 2026, expanding by market) can fill gaps on high-value prompts where organic citation is still contested. It runs on a relevance-weighted second-price auction, labelled "Sponsored" and shown separately from the AI's own answer — it supplements the organic strategy, it doesn't replace step 3.

How Suggesting.ai runs these five steps

We run this exact sequence for finance, legal and other regulated B2B clients, starting with a free 48-hour audit so you see steps 1 and 2 before committing to anything. From there we scope the content build (step 3), take on the weekly testing (step 4), and manage ChatGPT Ads and other LLM ad platforms as they open (step 5) — one accountable program instead of five disconnected vendors.

The reason we insist on running all five steps in order, rather than letting a client skip to step 3 because it feels like "the real work," is that we've seen how much budget gets wasted rewriting content for a site that's blocked at the crawler level, or optimizing for prompts nobody actually searches. Sequencing protects the budget as much as it protects the timeline.