How an AI Visibility Audit Finds the Revenue Search Is Losing

Where search revenue goes when AI answers first

Search revenue rarely disappears in one visible drop. It leaks. A buyer who once clicked three comparison results now reads an AI Overview or asks ChatGPT, gets a shortlist of three vendors and visits only those. If you are not on the list, the loss shows up as a slow fall in clicks while impressions hold, and as fewer demo requests that nobody can explain.

The leak takes three forms. You are absent from answers to unbranded buying questions. You are present but described wrongly, with old pricing, a missing feature or the wrong licence. Or you are mentioned, but the answer cites a third-party page that sends the click somewhere else. Each form needs a different fix, which is why the audit has to separate them.

Tracing the leak back to specific buyer questions

The audit starts from your money queries, not from AI. Pull the queries and pages that drove leads or sales over the past year, group them by intent and rewrite them as the questions a buyer would ask an assistant. Then run those questions on ChatGPT, Perplexity, Google AI features and Copilot and record the live answers.

For each question, note whether you appear, which competitors are named, which sources are cited and whether your facts are correct. Then check the causes on your side: whether AI citation crawlers can reach the relevant pages, including CDN and firewall rules that block bots even when robots.txt allows them; whether the pages state the answer plainly; and whether your entity data, meaning company name, products, locations and credentials, is consistent across your site and listings.

Putting a defensible number on the loss

A credible estimate uses your numbers, not industry averages. Take the query groups where AI answers now dominate and compare clicks year over year using Search Console's Web data, which already includes traffic from AI Overviews and AI Mode. Multiply the lost clicks by your historical conversion rate and average deal value for that page group.

Then add the opportunity side. Studies report that AI referrals convert several times better than classic organic visits; one 2026 study measured 15.9% against 1.76%. Treat that as a direction, not a forecast for your business. The honest output is a range per query group with the assumptions written down, so finance can challenge it and you can re-measure later.

Worked example: a broker losing its comparison traffic

Consider a forex broker whose best lead source was a set of comparison pages: lowest-spread brokers, brokers with Islamic accounts, brokers licensed in specific countries. Over a year, impressions on those pages held but clicks fell sharply, and account openings from organic search followed.

An audit of a situation like this would typically show AI answers to those questions naming four or five rivals, citing two affiliate review sites that do not list the broker, and stating the broker's minimum deposit from an outdated source. The fixes follow directly: earn accurate coverage on the cited review sites, restate account terms in plain text on the site and correct the stale data where it originates. This illustrates the method; it is not a client result.

The same pattern appears outside finance. A B2B software firm loses alternatives queries to review platforms; an industrial supplier loses specification questions to distributors whose catalogue pages state the answer more plainly. The audit method is identical; only the prompts and sources change.

Questions to ask before trusting a revenue estimate

• Which exact prompts were tested, on which engines, from which locations and on what dates?
• How many times was each prompt run, given that AI answers vary between runs?
• Was crawler access tested at the network level, not only read from robots.txt?
• Which of your own data sources fed the loss estimate, and what assumptions were used?
• Are fixes ranked by expected revenue impact and by effort?

If an auditor cannot answer these, the revenue figure is a sales number, not an analysis. A good estimate is also conservative about attribution: it should say which part of the decline is explained by AI answers and which part may come from seasonality, pricing or competitor spend.

What Suggesting.ai does with the findings

Suggesting.ai runs the first audit free and delivers it in 48 hours, with the prompts, answers, competitors, sources and a ranked fix list. From there we work on both sides of recovery. Organic GEO fixes access, content structure, entity data and third-party citations so AI engines can recommend you.

Where your category and market are eligible, ChatGPT Ads can restore visibility on high-intent conversations faster while organic work builds; ads are labelled Sponsored and shown separately from the answer. For forex brokers and other financial firms, OpenAI's policy generally rules out ads outside the US, so eligibility is confirmed in the audit, never promised. The point is to be the brand AI is suggesting when the buyer is ready.

Proving the revenue comes back

Re-run the same prompt set monthly and report mention rate, citation share and answer accuracy per query group. Segment AI traffic in analytics: ChatGPT search links carry utm_source=chatgpt.com, and Perplexity, Gemini and Copilot appear partly as referrers. Track leads and revenue from those segments against the baseline estimate. Because many AI answers are never clicked, also log branded search growth and how often prospects name an AI assistant as the place they first heard of you.

Set expectations with finance early. Recovery is rarely linear, AI answers shift between runs, and some lost clicks will not return because the question is now fully answered on the results page. The aim is to win the buyers who still need a vendor, and to be the name they remember when they do.