Five Gaps in Current AI Visibility Tools, and What to Ask Before You Buy

More than 300 AI visibility tools now exist, and almost all of them sell the same thing: a dashboard that counts how often your brand appears when software asks AI assistants a fixed list of questions. That number is cheap to produce, hard to trust, and rarely tied to revenue.

This piece sets out where current tools fall short. It draws on a feature review of 97 products from our full list of AI visibility tools and on published research into how AI answers behave. Much of that research comes from the vendors themselves, and we say so where it does. Current tools fall short in five areas: measurement, audience, revenue, action and agents.

Before these tools, marketers had no way to see whether AI assistants mentioned their brand at all. The number on the dashboard is still an estimate, and buyers should know how good an estimate it is. We wrote earlier about the deepest limit, the causality problem: the tools show what happened and leave the cause unexplained. This piece looks at everything around it.

What the tools do today

The category has converged on one product. Nearly every tool runs a list of prompts, counts brand mentions, compares you with competitors and lists the sources cited. Far fewer can say who was asking, what it earned, or what AI agents did next.

Capability	Share of 97 products that state it
Prompt tracking	98%
Share of voice vs competitors	91%
Citation and source analysis	82%
Content optimisation	77%
Sentiment	69%
API and integrations	67%
White-label reporting	56%
Technical audit (llms.txt, schema)	53%
Agents and automation	52%
Country and language tracking	49%
Prompt discovery and volume	48%
AI traffic attribution	39%
AI crawler analytics	30%
Persona or audience segmentation	28%
Shopping visibility	23%

Feature review of 97 products (67 dedicated tools and 30 modules inside larger suites) from vendor websites, 24 to 27 October 2026. A capability counts only when the vendor states it.

The capabilities most tools share describe the same activity: counting appearances in answers to generic questions. The rarer ones, in bold, are where the five gaps begin.

Gap 1: The core number is unstable

AI answers change almost every time you ask. In a January 2026 study with Gumshoe, SparkToro had about 600 volunteers run 12 prompts through ChatGPT, Claude and Google's AI 2,961 times. The same list of brands came back in under 1% of runs, and in the same order in under 0.1%. Rand Fishkin's verdict: visibility measured across many prompts, each run 60 to 100 times, is a reasonable metric, but "any tool that gives a 'ranking position in AI' is full of baloney". Profound's own research says up to 90% of cited sources can change over time.

Sampling is where tools differ most. Profound runs each prompt once a day per platform, and its own test of 753 prompts found that daily readings landed within about two percentage points of a ten-runs-a-day setup (Profound). That holds when you average many prompts over weeks. It does not hold for one prompt on one day, which is what many dashboards show.

Four further problems sit on top of that:

API answers differ from the app. Surfer compared 13,779 answers and found only 15–32% overlap between the brands the API returns and those users see in the app. Brandlight matched 834 ChatGPT prompts and found 11% brand overlap. An October 2026 academic paper found that Google results plus the OpenAI API recover only about a quarter of the pages ChatGPT cites. A tool that reads the API to save cost is measuring an answer nobody sees.
Every user sees a different answer. ChatGPT rewrites prompts using memory and location, and Gemini's personal context is now on by default. Lily Ray found tracker results almost never matched what she saw herself.
"AI search volume" is mostly modelled. No AI company publishes what people ask. Profound scales third-party panel data up to the global user base and acknowledges the risk that its panel "lacks representativeness" (Profound). Ahrefs says there is "no significant, reliable source of data" on how people prompt AI and derives its estimates from Google search volume instead (Ahrefs). Conductor points out that even samples of tens of millions are under 1% of daily prompts. None of these vendors publish error margins.
Model updates move the numbers overnight. SISTRIX tracked 3.8 million German-language ChatGPT answers: citations normally varied 1–2% a day, but jumped to 47% around the GPT-5.5 release in May 2026. In a smaller test, Writesonic found the new model cited brand websites half as often. Trackers even disagree about the same event: Semrush measured ChatGPT's 2025 Reddit drop as a fall from about 60% of responses to about 10%, while other tools reported a far lower starting point. Few dashboards mark model releases, so teams react to drops they didn't cause.

Meanwhile the free baseline is improving. Bing Webmaster Tools added an AI performance report in February 2026, and Google Search Console began showing AI Overview and AI Mode impressions in June 2026. Basic visibility counts are becoming free.

A reliable reading needs appearance rates measured over many runs, from the consumer app, with confidence intervals and model changes marked. Few tools offer that today: IQRush puts intervals on every metric, and Evertune samples each prompt up to 100 times. Most report a single number with no sense of how much it would move tomorrow.

Gap 2: The buyer is missing

Most trackers ask AI assistants a neutral question, "best CRM for small businesses", as if no one in particular were asking. Real buyers bring context: their role, budget, sector and what they asked a minute ago. ChatGPT has drawn on past conversations to personalise answers since April 2025 for paid users and June 2025 for logged-in free users (BleepingComputer), and its shopping research feature uses that memory when it recommends products (The Decoder).

That context changes the answer. Berlin agency Radyant ran 17,929 chats and found that adding a role to the prompt moved a brand's visibility by up to 24 percentage points, with Google's AI Overviews about four times more sensitive to it than ChatGPT. A May 2026 preprint found that category leaders stay about 80% stable across personas, but mid-market brands can see up to 75% of their recommendations replaced. People also phrase the same need very differently: when 142 people wrote their own prompt for the same headphone purchase, the prompts had an average semantic similarity of just 0.081 (SparkToro).

Only 28% of the 97 products we reviewed say they segment results by persona or audience. Just seven tools are built around it (Chatoptic, Gumshoe, Parsnipp, RivalSee, Am I On AI, QueryBurst and Rankfor.AI), and most of those target small businesses and consumer brands.

The gap is widest in B2B, where a purchase involves several people asking different questions:

The buying committee is invisible. A CFO asks about cost and risk, an IT lead about integration and security, an end user about ease of use. A tool that tracks one generic prompt set can't tell you which of them your brand is losing.
B2B buyers already start here. In Gartner's 2025 buyer survey, more buyers discovered software through generative AI (9%) than through search engines (8%). G2 reports that half of B2B buyers now start in an AI chatbot.
The questions are niche. Gartner's Market Guide lists coverage of "complex, low-volume B2B queries" as a mandatory feature. Volume-based prompt discovery, built from consumer search data, finds the wrong questions.
Buying happens over a conversation. Buyers refine, compare and push back over several turns. Only a couple of tools (Parsnipp and Chatoptic) simulate multi-turn journeys to find where a brand drops out of consideration.

For a market leader, an average visibility score may be close enough. For a challenger that wins with one type of buyer and loses with another, the average hides exactly what matters.

A useful report breaks results down by the people who actually buy, including each role in a B2B buying committee, measured across a whole conversation rather than one prompt.

Gap 3: Visibility isn't revenue

A rising visibility score is only useful if it leads to customers, and most tools can't show that it does. Only 39% of the 97 products we reviewed claim any AI traffic attribution, and in most cases that means counting referral visits in Google Analytics.

Referral counts badly understate AI's effect:

Clicks are falling. When Google shows an AI Overview, users click a traditional result on 8% of visits against 15% without one, and click a link inside the summary on just 1% (Pew Research Center).
Direct AI traffic is small. Studies put AI assistants at between 0.3% and about 1% of website traffic, against roughly 25% from organic search.
Much of it arrives unlabelled. One analytics vendor found that 70.6% of AI visits carried no referrer and showed up as "direct" traffic (via David Rosam). Google's new generative AI report in Search Console shows impressions but no clicks.
But it converts. Adobe measured US retail traffic from AI up 393% year on year in Q1 2026, with those visitors converting 42–54% better than others by spring.
Most of the effect arrives by another route. Similarweb found brands that ChatGPT recommends are 2.5 times more likely to get a visit within a week, but 56% of those visits come through branded search, not an AI referral link.

The analysts agree this is the frontier. Gartner lists revenue attribution as a future capability, and Forrester names linking AEO to business outcomes as a main way vendors will win. Investors are backing it: Petra Labs raised a $5.2M seed in July 2026 specifically to build attribution models for AI search.

What's missing: a model that connects visibility changes to branded search, direct visits, pipeline and revenue, so a marketing team can defend the budget.

Gap 4: Diagnosis and fixes

Most tools now suggest fixes: 77% of the 97 products offer content recommendations or generation, and 52% offer some kind of agent or automation. But most of those fixes are on-site content edits, and AirOps estimates that about 85% of brand mentions in AI answers come from third-party pages: reviews, press, Reddit, comparison sites. Forrester now describes AEO as cross-functional brand work spanning PR, social and community, not an SEO task. A content checklist misses most of the levers.

The judgement that connects the data to a plan (which buyers you lose, why, and what to change first) sits with agencies. Agencies price it on a different scale:

Offer	Products with a published price	Lowest	Median	Highest
AI visibility software	87	$6.89/mo	about $91/mo	$599/mo
GEO agencies	7	$1,500/mo	$6,000/mo	$10,000/mo

Entry prices from vendor pricing pages and agencies' published minimums, checked 24 to 27 October 2026. Euros converted at 1.15 and pounds at 1.27; yen-priced and custom-priced offers excluded. Omniscient Digital's $120,000 a year shown as $10,000 a month.

Among recurring monthly offers, none with a published price sits between $599 and $1,500 a month. One-off audits cover more of that range: agencies in our full list publish audits such as Arobis AI's at $399 and Visiblie's from $1,000, up to five-figure engagements such as Re:signal's £10,000 audit. Persona tools such as Gumshoe and RivalSee give away a one-off scan as a free trial, and we found no tool selling a paid, self-serve diagnosis.

The missing offer is a decision-grade diagnosis priced like software and as specific as an agency's, covering off-site sources as well as your own pages.

Gap 5: AI is starting to act, and the tools only read

AI visibility tools measure what assistants say. Increasingly, assistants also act: they crawl sites on demand, pull product data from feeds, compare offers and complete purchases. Only 30% of the 97 products we reviewed track AI crawlers, and 23% track shopping visibility.

Buying inside AI assistants arrived in a year. ChatGPT launched Instant Checkout in September 2025 (Fintech News) and reportedly scaled it back in March 2026 (EcomCrew). Perplexity and PayPal launched Instant Buy in November 2025, and Google announced agentic checkout in AI Mode and Gemini in January 2026. Agents also choose differently from people: a study of AI shopping agents found strong position biases, and that model updates "can drastically reshuffle market shares" (arXiv).

Three changes in 2026 make this layer hard to ignore:

Access is becoming a setting you can get wrong. From 15 September 2026, Cloudflare blocks AI agent and training crawlers by default on ad-supported pages for new sites and free plans, according to secondary reports. Pay per crawl, TollBit and AWS now let sites charge bots. A brand can disappear from answers through a CDN setting no visibility dashboard shows.
Shopping answers now run on feeds. Profound's data shows the share of ChatGPT Shopping picks drawn from merchant feeds jumped from 8% to 62% on 10 July 2026. Product data quality decides those answers. Shopify, Google and OpenAI are competing over the protocols.
Paid placement is arriving. Amazon began charging for Sponsored Prompts inside its shopping assistant in March 2026, and Profound, Evertune and Brandlight now sell ads inside AI answers. Organic visibility will soon need to be read alongside paid.

Gartner already warns that brands failing agent checks risk becoming "transactionally invisible", and lists monitoring of agentic transactions as a future requirement. None of the vendor methods we reviewed measure what an agent picks, compares or buys; they record what an assistant says.

Brands need one view of what AI says, what AI crawlers can reach, what product data agents use, and what is paid versus earned.

What the next generation of tools needs to do

Gap	What most tools do today	What's needed
Measurement	One run per prompt, often via the API	Many runs from the consumer app, reported with confidence ranges and model changes marked
Audience	Generic prompts asked by no one in particular	Results by buyer persona and, in B2B, by each role in the buying committee, across multi-turn conversations
Revenue	Referral visits in GA4	Modelled lift on branded search, pipeline and revenue
Action	A dashboard and a list of suggestions	A prioritised diagnosis that says what to change, where and why, at a price between software and agency
Agents	Answers only	Crawler access, product feeds, agent actions and paid placement in one view

Eight questions to ask before you buy

Until tools close these gaps, buyers should ask vendors directly:

How many times is each prompt run, and do you show a confidence range or a single number?
Do you collect answers from the consumer apps or from the API, and for which engines?
Can you show my visibility for different buyer types, and across multi-turn conversations?
Where do your prompt volume figures come from, and how large is the sample?
Do you flag model and search-system changes, and can I compare before and after?
How do you connect a change in visibility to visits, leads or revenue?
Do your recommendations cover third-party sources, or only my own pages?
Do you track AI crawler access, product feeds and what AI agents select?

And spend accordingly:

Use the free baseline first. Google Search Console and Bing Webmaster Tools now report AI citations at no cost.
Buy tracking cheaply. Basic monitoring is a commodity, and entry plans at $30–$100 a month cover the core metrics.
Spend on understanding. The value lies in knowing which buyers you lose, in which conversations, and what would change that.
Check your crawler and feed settings. A blocked bot or a weak product feed can cost more visibility than any content change gains.

From measuring AI visibility to improving it

Knowing that a competitor appears more often in AI answers raises the questions that matter: which buyers choose that competitor, in which conversations your brand drops out, and what supports the recommendation. A visibility score can't answer them on its own.

That is the problem ZIO is built around. It is designed to use synthetic buyer personas to show which audiences AI assistants recommend your brand to, and which they don't, across ChatGPT, Gemini and Claude. If you are already tracking AI visibility and aren't sure what the results mean, or need help choosing a tool, talk to the ZIO team. To compare what exists today, see our directory of GEO and AI visibility tools.

Method and sources

Feature figures come from a review of 97 AI visibility products (67 dedicated tools and 30 modules inside larger suites), scored on 15 capabilities from each vendor's own website between 24 and 27 October 2026. They reflect what vendors claim, not hands-on tests: a capability counts only when the vendor states it, and a missing description is not proof a feature is absent. Prices were checked on vendors' pricing pages on the same dates. Several studies cited here were published by companies that sell AI visibility tools, including Profound, Ahrefs, Surfer, Brandlight, Writesonic, Semrush, Conductor and Similarweb.

Disclosure: ZIO builds persona intelligence for AI search, so we work in this market and have a view on it.