AI Answers Are Swallowing the Click, and the Tools Built to Track Them Are Changing Shape

The numbers that changed the conversation

Start with traffic. In July 2025 the Pew Research Center published a study built on the browsing data of about 900 US adults, covering 68,879 Google searches made in March of that year. When the results page showed an AI Overview, people clicked an ordinary result in 8% of visits. Without one, the figure was 15%. Links inside the AI summary itself? Clicked in roughly 1% of visits.

Ahrefs saw the same thing from the publisher’s side. Looking at 300,000 keywords, it found the top‑ranking page got about 34.5% fewer clicks when an AI Overview sat above it.

Then there’s reach. OpenAI says ChatGPT has around 800 million weekly users. Google has said AI Overviews reach more than 2 billion people a month. Back in February 2024, Gartner predicted that traditional search engine volume would drop 25% by 2026 as people moved to chatbots and virtual agents. Whether that exact figure lands is anyone’s guess. Nobody in search marketing is betting against the direction, though.

Put those numbers side by side and the thing companies compete for changes. Ranking third on a results page is worth less if the buyer never scrolls the page. What counts is being one of the three or four sources an AI answer quotes, and being described accurately when it does.

Measuring was the easy half

The first wave of tools went after measurement, and did it well. Profound, probably the best‑known name in the category, raised a $35 million Series B led by Sequoia in August 2025, bringing its total funding to $58.5 million. Its platform shows how brands appear across ChatGPT, Perplexity, Gemini and other engines, and it has since added content generation and marketing agents on top. Dozens of smaller trackers followed. Semrush and Ahrefs both bolted AI visibility modules onto their suites.

These products tell you which prompts you’re missing from. They count competitor mentions. Some list the domains the engines cite most often for your topic, which is genuinely useful.

What they leave on your desk is the part that moves the result. The report lands, it’s correct, and the to‑do list it implies (find the sites the engines trust, pitch their editors, negotiate, write the piece, get it published, check that it’s indexed) belongs to nobody in particular. For a five‑person marketing team, that list is a quarter of work.

Why the engines care who mentions you

Research gives a decent idea of why the list looks like that. A 2023 paper on “generative engine optimization” by researchers from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi (Aggarwal et al., presented at KDD 2024) tested how edits to source content changed its visibility in AI‑generated answers. Adding citations, quotations and statistics lifted visibility by up to 40% in their benchmark. Keyword stuffing barely helped.

Underneath, the engines still run on search. ChatGPT’s browsing has leaned heavily on Bing’s index. Perplexity crawls and ranks its own results. AI Overviews draw on Google’s ranking systems. So the old signals are still there: authority, topical relevance, other trusted sites referencing you.

The difference is that a language model reads the words around a brand name. A mention in a respected trade publication that explains what a product does, who it’s for, how it compares, teaches the model something. A dozen directory listings that only print the name teach it very little.

All of which makes for slow, manual, unglamorous work.

From dashboard to done‑for‑you

That gap is where the newer services are setting up. Instead of selling the diagnosis, they sell the fix.

One example is the generative engine optimization service RankCited, which reverses the usual order. A brand signs up with its domain and gets a free visibility report first: the questions buyers in its category ask ChatGPT, Perplexity and Google’s AI, which answers name competitors instead, and which sources those answers rely on. The paid part is the work itself. Each plan buys a month of capacity. RankCited’s AI agents turn the gaps into a placement plan, a human strategist approves it, and the team handles the outreach, the editorial back‑and‑forth and publication. Then it re‑measures across ChatGPT, Perplexity, Gemini and Google’s AI answers.

According to the company, placements come in three shapes. A niche industry publication the models already quote on the topic. The right vertical or section inside a large outlet. Or a feature commissioned so that the surrounding text explains the brand properly. Publisher names stay private until a piece goes live, partly so a client’s pipeline doesn’t become a public list a competitor can copy.

It’s a different bet from the dashboard players. Software margins beat service margins, and investors know it. Still, mid‑market buyers seem to be asking for outcomes rather than another analytics login, and plenty of agencies now sell “GEO” retainers chasing the same demand. Which model wins probably depends on the buyer. A company with an in‑house content and PR team can do a lot with a good tracker. One without it mostly ends up with a well‑documented problem.

What to ask before paying anyone

Tracker, agency or managed service, a handful of questions separate the serious options from the noise:

• Which engines get measured, and how often? ChatGPT, Perplexity, Gemini, Google’s AI Overviews, Claude and Copilot answer differently. One engine checked once a month misses most of the picture.
• Can the engines actually read the placements? A sponsored post parked in a noindexed section, or on a site that blocks crawlers like OAI‑SearchBot and PerplexityBot, won’t get cited. Ask to see.
• Is the context right? Fifty mentions on unrelated blogs say little about what you do. Five in the right trade publications can change how a model describes you.
• How are results reported? AI answers shift from day to day. Anyone promising a fixed spot inside ChatGPT is guessing.
• Who does the work? If the deliverable is a list of tasks for your own team, price it like software.

Still early

Very little of this is settled. Google keeps expanding AI Mode to new markets, OpenAI keeps adjusting how ChatGPT search picks its sources, and the research on what makes a page “citable” is barely two years old. The tools will keep changing with it.

One shift does look permanent. Finding out that you’re invisible in AI answers has become cheap. Fixing it hasn’t, and the market is only starting to put a price on that difference.