AI Visibility Tracking Tools in the UK: Twelve Vendor Pages Fetched, Five Surfaces Named in the Markup, and Two With No Price In It

A British buyer trying to build a price comparison of AI visibility trackers is working from pages that often do not contain a price. On 16 October 2026 we requested twelve pages belonging to six vendors in this category, one request each, no JavaScript run. Nine returned HTTP 200. Of the five pricing pages that answered, two carried no money figure of any currency in the visible first response, and across all twelve responses every figure we could extract was in dollars. Not one pound figure appeared in the set. So the comparison worth building in the UK is not the price table. It is the five columns a vendor page or a two week trial can confirm.

What was fetched, and how to repeat it

The method is deliberately dull so that you can run it on the morning you are buying. Each URL was requested once, on 16 October 2026, with a desktop Chrome user agent string and no cookies. Nothing was rendered: scripts, styles and tags were stripped and what remained was counted as visible text. A money figure counts as present when a currency mark or code sits beside a number in that text. The surface count asks how many of ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews are named in the same response.

page requested	status	visible words in first response	money figure present	surfaces named, of five
peec.ai	200	3,858	no	5
peec.ai/pricing	200	3,454	no	5
tryprofound.com	200	2,175	yes	5
tryprofound.com/pricing	200	1,302	no	5
ahrefs.com/brand-radar	200	2,224	yes	5
ahrefs.com/pricing	200	2,037	yes	1
semrush.com/ai-toolkit/	404	656	no	3
semrush.com/prices, which redirected	200	1,368	yes	0
scrunch.com, reached from scrunchai.com	200	5,244	yes	2
scrunch.com/pricing	200	1,058	yes	5
otterly.ai	403	15	no	0
otterly.ai/pricing/	403	15	no	0

Two rows need reading rather than quoting. The otterly.ai pair is out of every count above the table, because a 403 under one reading is not an access decision: that host answered 200 with its full article hours apart on an earlier run of ours, and an explicit HTTP 429 turned up elsewhere in the same set. The rule that reading produced sits in our bot access audit where only two of four apparent blocks survived a second reading. The semrush.com/ai-toolkit/ row is a 404 on a path we assembled, which is our error.

The finding a British buyer should take from it

There was no pound figure to read. Every extractable money figure across the twelve responses was in dollars, so a pound number you are shown for any of these tools is somebody's conversion, worth reading only with the rate and the date of that rate beside it. The VAT basis sits on top of that, and it is settled first rather than last: see what a UK buyer should settle before the pound price.

Two of the five pricing pages that answered had no price in the document at all. That is not a criticism of either product. A figure that arrives after a script runs is invisible to anything reading bytes, and plenty of things read bytes. The price on those two is a conversation rather than a column.

The surfaces were easier to read than the prices. Six of the twelve responses named all five surfaces, and both pages with no money figure were among them. So the engine list usually comes off the page while the cost often does not. Naming a surface in marketing is not the same as it sitting inside the tier you were quoted, so the list starts a question rather than ending one.

Our own page failed the same test this morning, and the cause is the useful part

We sell a tracker in this category, so the measurement includes us, and on the first reading of the day we failed it. Fetched as GPTBot, our pricing page served rupee figures and no dollar, pound or euro figure. Re-fetched the same day after a fix, the visible first response carries dollar, pound, euro and rupee figures together.

The cause is worth more than the fix. The region detector was never broken. A browser sends a country header from the network edge, a time zone or a language preference and gets its own ladder. A crawler sends none of those, so the detector reached its last line, which says that when we cannot tell, show rupees. That default is deliberate: it protects an Indian buyer whose Android handset ships with a British locale from a ladder more than twice his own. The defect was that the other currencies sat behind a browser toggle, never in the document.

And our own check could not see it, because it compares what a crawler receives with what a browser receives, and from our network both received rupees. A check can pass because it is blind in the same direction as the thing it checks. The replacement asserts the arithmetic instead: three currencies in the table, India in rupees, the United States in dollars, the United Kingdom in pounds.

The five columns that survive a check

• Engines inside your exact tier, each of the five written out individually, with the cost of any that are extra. Two quotes can be compared on a list and never on a phrase, so make the supplier produce the list.
• The unit the allowance is counted in. A tracked question, an engine, a recorded answer and a domain are four different meters, and two honest quotes for identical work can differ by a multiple.
• How often the set is re-run. Every step up the schedule puts more answers on file, and the answers on file are frequently what the supplier is billing.
• Raw answer export. Treat it as required. Without the answers you cannot audit the number you are reporting to somebody who was not in the demo.
• Whether a rate arrives with a range, and whether the method is published. Twenty answers produce a count, and a count dressed as a percentage with no interval beside it reads larger than it is. Ahrefs links a Brand Radar methodology post from the page we read on 16 October 2026, which is the shape of answer to ask every vendor for.

The free tier belongs in the same grid, scored on coverage rather than on the word free: how many questions, how many engines, whether it needs an account, whether it refills.

Why the price column stays empty here

This site records prices and does not reprint them, our competitors' included, and the position is not a style choice. It rests on a reading taken on 17 September 2026 in which an engine checked a comparison page's competitor figures, agreed they were right, and discounted the page anyway because nothing beside them said where or when they came from. Correct was not enough. Undated was fatal. The eleven column field list for building your own dated row is in AI visibility tools compared, and a cell needing a sales call stays empty until the call happens.

What a British buyer asks when comparing AI visibility trackers

Did any of the twelve pages show a figure in pounds?

Not one, on 16 October 2026. Dollars were the only unit anywhere in the set, which means somebody has done a conversion before a sterling number reaches you. Ask who did it, at what rate, and on what date.

Can I tell from a vendor's own page whether Claude is included?

You can usually read the surface list off the page, because six of the twelve responses named all five. Whether Claude sits in the tier you were quoted is a separate question and it belongs on the quote, written out surface by surface with the add on cost of any that are extra.

Why will you not print the competitors' prices?

A figure copied into a table here stops being true the next time somebody else changes a plan, and we would have published a wrong statement about another company without noticing. Handing you the worksheet outlasts handing you the table.

How do I run this check on a vendor myself?

Request the pricing page with a plain HTTP client and read what comes back. If the figure is absent from that response, it is absent from the document an automated reader sees as well, and the purchase will involve a call.

Is a free tier enough to decide on?

To find out whether you have a problem, yes. To defend a figure in front of a finance director, the four things you are buying are a question list that does not move, one column per surface, a schedule you keep, and the answers themselves on disk. Redo this fetch check monthly: our own page changed within hours of being measured.

The question behind all of this

Ask ChatGPT or Gemini what one of these trackers costs in pounds and an answer arrives, tidy and confident, assembled from pages that mostly do not contain the figure. What those two engines actually reach for is published as 597 readings across ChatGPT and Gemini in three markets, and the British half of our later run cited our own domain in 3 of 210 ChatGPT readings and 0 of 210 on Gemini. We build the tool that takes that reading, which is why this page includes the row where we failed. If you sell in Britain, the practical move is the one we had to make: put the pound figure in the document, not in the script.