Is AI visibility worth paying for? A business case from 2026 traffic, conversion and price data

What does a finance team need to see?

A finance team needs three things on one page: how many buyers AI answers reach, what those buyers are worth, and what the work costs. Most AI visibility pitches supply the first and skip the other two. This post puts the published 2026 evidence for all three in one place and turns it into a template you can fill in.

The underlying research belongs to other people. Similarweb traced what happens after ChatGPT recommends a brand, Adobe measured how AI-referred shoppers behave on US retail sites, and Seer Interactive compared clicks for pages cited and not cited in Google's AI Overviews.¹ What this post adds is the arrangement: each study placed on the line of the business case it can support, and a three-scenario model with the assumptions named.

Finance teams now sit in more software and services decisions, and they ask harder questions about units, terms and exits. Our guide to getting an AI purchase past the CFO covers the approval process. This post covers the numbers that go into it.

A disclosure first. Sigzen AI sells AI visibility audits and programmes at published prices, so we have an interest in the answer. We have tried to show where the answer is no. The model works the same whoever you buy from, and if you do the work in-house.

How much traffic does AI search actually send?

Not much, if you count clicks. In SE Ranking's study of 101,574 websites in 250 countries, AI search sent 0.32% of all visits in January–April 2026.² For most sites that line is a rounding error next to organic search.

The absolute numbers are larger than the share suggests. Similarweb counted an average of 770.7 million AI referral visits a month worldwide between June 2025 and May 2026.³ The audience behind those visits is larger still: at its DevDay on 29 September 2026, OpenAI reported a collective 1.2 billion weekly users, a figure the company published about itself.⁴

Our view: a business case built on referral clicks alone will usually come out small, and it should. A CFO who sees 0.32% will ask why the budget line is not a similar share of the search budget.² The answer has to come from the value of AI-referred visits and from influence on other channels, which the next two sections cover. If neither holds for your buyers, the small share is the honest result.

The volume side

A small share of clicks from a very large audience

0.32%of website visits came from AI searchSE Ranking, Jan–Apr 2026 ²

770.7MAI referral visits a month, worldwideSimilarweb, Jun 2025–May 2026 ³

1.2Bweekly users, as reported by OpenAIOpenAI, Sep 2026 ⁴

Do AI-referred visitors convert better?

On US retail sites they do, and the gap widened during 2026. Adobe reported that AI-referred traffic to US retail sites rose 393% year on year in Q1 2026. In March 2026 those visits converted 42% better than non-AI traffic and earned 37% more revenue per visit.⁵

By July 2026 the premium was larger. Adobe found AI referrals to US retail sites up 62% year on year that month, with AI visits converting 60% better than other traffic and adding items to the cart 28% more often.⁶ Growth slowed while conversion improved, which is how a channel behaves as it matures.

Two cautions before this goes into a spreadsheet. The data covers US retail only, and we have not found a comparable published series for other countries or for B2B. And a conversion premium is partly selection: shoppers who arrive from an AI answer have often done their comparison already, so they would convert well from any channel.

Our view: use the premium as evidence that an AI-referred visit is worth more than an average visit. Do not use it as the uplift your own site will see.

AI-referred visits compared with other traffic, US retail sites

Conversion premium, July 202660%

Conversion premium, March 202642%

Revenue per visit premium, March 202637%

Add-to-cart premium, July 202628%

Where most of the value shows up: influence, not clicks

Most of the value of being recommended arrives through other channels, mainly search. That is the strongest finding for a business case, and the easiest to miss in an analytics report.

In Similarweb's US desktop panel (July–December 2025), brands recommended by ChatGPT were 2.5 times more likely to be visited within seven days. Of those visits, 55.9% arrived through search, and only a small share as visible AI referrals.⁷ The study covered three brand pairs in finance, travel and beauty, so read the direction as solid and the size as specific to those brands.

Google's AI answers cut the other way for pages they leave out. Ahrefs found the top-ranking page's click-through rate was 58% lower when an AI Overview appeared, comparing December 2023 with December 2025 across 300,000 keywords on desktop (geography not stated).⁸ In Seer Interactive's panel of 53 brands (January 2025–February 2026), pages cited in an AI Overview earned about 120% more organic clicks per impression than when they were not cited.⁹

Taken together, the studies describe one mechanism. AI answers decide which brands a buyer meets, and the buyer then searches for those brands by name. A brand left out of the answer also loses clicks it used to win from rankings. Our view: this is the core of the business case, and it is why AI referral counts belong low in a board report.

What B2B buyers report

B2B evidence comes from surveys rather than traffic panels. In G2's survey of 1,076 B2B decision makers in North America, EMEA and APAC (March 2026), 51% said they start software research in an AI chatbot more often than in Google.¹⁰ In the same survey, 69% said an AI chatbot had surfaced information that led them to choose a different vendor than they expected.¹¹

People who use AI also check more places. In Gartner's survey of 1,004 US consumers (August 2026), shoppers using AI for inspiration consulted 4.48 sources on average, against 3.24 for those who did not.¹² AI adds a step to the journey rather than replacing the others. The answer has to agree with your site, your listings and your reviews, or the extra checking works against you.

What each study can and cannot carry

Each study supports one line of a business case, and none supports all of it. The table maps them, so finance can see which source holds up which assumption.

Our reading of the sources cited in this post.

The weakest link in most decks is the step that turns influence into money. No published study gives a revenue figure per mention or per citation, and any vendor that quotes one is extrapolating. You have to supply that step from your own funnel.

Our view: that is a feature of the model, not a flaw. A case built on your deal size, your margin and your buyers' answers to "how did you find us" survives a finance review. A case built on a borrowed multiplier does not.

What does AI visibility work cost in 2026?

Published prices put agency retainers in the low thousands of US dollars a month and tools in the tens. In our dataset of published GEO prices, seen on 5–6 October 2026, 34 of 41 agencies listed a price, alongside 25 tools.¹³ Across the 13 listed agencies that price in US dollars, the median monthly entry price was $3,400. Across the 16 tools with a monthly US-dollar entry tier, it was $79.¹³

Entry prices are floors. They rarely include off-site citation work, and tools measure without fixing anything. Our comparison of published GEO prices sets out what each tier includes and how agency, tool and in-house options compare.

Our own prices are on the pricing page.¹⁴ The AI Visibility Audit is $3,500 (₹1.5L), one-time, read out in ten working days: 150 buyer prompts across four engines, each run on three separate days. The AEO/GEO sprint is from $8,000 (₹3L) over six weeks. Programmes run from $3,500 (₹1L) a month on a six-month minimum, in Core, Growth and Scale tiers.

Add your own team's time to any quote. Someone has to approve page changes, answer fact requests from the agency and read the reports. Internal time is the line most business cases leave out, and it is the first one finance asks about.

A three-scenario model: defend, hold and grow

Model three spend levels and ask how many extra customers each needs to break even. That turns a vague question about visibility into a narrow one about deals or orders, which your sales and finance teams can judge.

The formula is short. Break-even customers equal the year-one cost, including internal time, divided by the gross profit from one new customer in that year. Then ask whether AI answers could plausibly move that many customers, given the evidence above and what your own buyers say.

The three scenarios

Defend means measuring only: a baseline across engines, a register of wrong facts and an alert when a competitor replaces you. Hold adds a fix sprint so pages, entity data and listings say the same thing. Grow adds a programme of off-site citation work and new answer-shaped content for six months or more.

Each level answers a different worry. Defend answers "are we being misdescribed?", hold answers "are we losing ground we had?", and grow answers "can we take share from a competitor?". A finance team can approve the first without believing the third.

Three spend levels for one business case

Defend

• Baseline across four engines
• Wrong-fact register
• Competitor alerts
• Breaks even on a fraction of one deal

Hold

• Everything in defend
• Pages answer first
• Entity data and listings aligned
• Breaks even on about one deal

Grow

• Everything in hold
• Off-site citation work
• New content each month
• Needs several deals a year

Read the two examples side by side. A high-value B2B sale breaks even on a handful of deals, so the debate is about whether AI answers touch the shortlist at all. A retail brand needs hundreds of orders, so the debate is about volume and how much of the search lift you can credit.

When does the business case fail?

The case fails when your buyers do not use AI to choose, when you cannot set a baseline, or when the break-even count is implausible. Say so in the deck, because a finance team trusts a case more when it names its own failure points.

• Your buyers do not ask AI. Ask your last 20 customers how they built their shortlist. If nobody mentions an AI assistant, buy defend at most.
• You cannot measure a starting point. Without a baseline across engines, any later change is an anecdote, and finance will treat it as one.
• The arithmetic does not close. When grow needs more than a few in every hundred customers to come through AI-shaped journeys, it is a bet, not a business case.
• The problem sits elsewhere. If a marketplace listing or a review site states your prices wrongly, fixing that source is cheaper than any programme.

Our view: most mid-sized companies should start at defend or hold and earn their way to grow. Ask a vendor, including us, to show the break-even count for each level before you sign anything longer than an audit.

How do you prove the case afterwards?

Agree the measures before any work starts: share of answer across engines, the AI referral line, branded search and self-reported attribution. Then compare each against the baseline at three and six months.

Share of answer needs a sound method, because answers change from run to run. Ours runs each buyer prompt on ChatGPT, Gemini, Google AI Overviews and Google AI Mode, through a licensed AI-answer data provider. Each prompt runs on three separate days, and results carry confidence intervals.

Then connect the answers to revenue. Add a free-text "how did you hear about us" field to forms and sales calls, and tag the answers in your CRM. Our guide to proving AI visibility in pipeline covers the field, the tags and the report that goes to the board.

Build and test the case in five steps

1. Ask your buyers

   Twenty recent customers: did an AI assistant shape the shortlist?

2. Set a baseline

   Share of answer across four engines, each prompt run on three days.

3. Price three levels

   Defend, hold and grow, with internal time in each.

4. Count break-even customers

   Year-one cost divided by gross profit per new customer.

5. Review at six months

   Share of answer, branded search, AI referrals and self-reported source.

If the six-month review shows share of answer up and nothing downstream, stop at hold. If it shows neither, stop. Our view: a programme that cannot show movement in its own measures by month six has not earned month seven.