An AI SEO tool is software that uses language or machine learning models to do part of the SEO work: writing briefs and drafts, grouping keywords by intent, generating titles and structured data, or measuring how often a brand is cited in the answers of ChatGPT, Perplexity and Google's AI Overviews.
Two jobs that have little in common live under the same label, and on page one of Google they almost always end up in the same list. Before looking at prices it is worth separating them, because that is where the budget gets decided.
Two different categories sold under the same name
The first category is AI inside SEO tools that already existed. Semrush, Ahrefs, Surfer SEO, SE Ranking and Frase have added a language model on top of their data: the task was already in the software, the model makes it faster. Here AI is a feature, not the product.
The second category measures a surface that did not exist before: how often you appear in generated answers. Otterly.AI, Profound, Peec AI, Rankscale and Ahrefs Brand Radar query the assistants at regular intervals and count citations and mentions. They do not optimise anything, they count.
Then there are general-purpose LLMs used by hand: ChatGPT, Claude, Gemini, Perplexity, NotebookLM. They are not SEO tools and have no search database behind them. The distinction sounds pedantic until you try to use them as if they were.
Further reading: the map of the classic suites, with prices, is in best SEO tools: how many you really need and what they cost.
Eight UK keywords, two models, one Keyword Planner
On 26 September 2026 we asked two models to estimate the average monthly Google search volume in the United Kingdom for eight keywords from our industry. Same prompt, web search switched off, no extra context: gpt-4.1-mini (version 2025-04-14) and gemini-2.5-flash. The same day we pulled the real volumes from Google Ads data for the UK in English, average of the last twelve months.
| Keyword | Real volume (Google Ads, UK) | gpt-4.1-mini estimate | gemini-2.5-flash estimate |
|---|---|---|---|
| ai tools | 9,900 | 12,000 | 135,000 |
| best seo tools | 1,300 | 8,000 | 6,600 |
| keyword research | 4,400 | 5,500 | 14,800 |
| seo audit | 1,300 | 4,000 | 4,400 |
| ecommerce seo | 720 | 3,200 | 3,600 |
| local seo | 1,300 | 3,800 | 5,400 |
| link building | 880 | 6,000 | 4,400 |
| core web vitals | 1,300 | 2,500 | 3,600 |
Two estimates out of sixteen fall within 30% of the real value. The other fourteen do not, and the gaps are large.
The worst case is "ai tools": 9,900 searches a month in the UK. gemini-2.5-flash said 135,000, nearly fourteen times as many. gpt-4.1-mini put "link building" at 6,000 against a real 880, almost seven times as many, and ranked it above "keyword research", which actually has five times its demand. Anyone who built a content plan on these estimates would have put the effort in the wrong order.
The detail that makes the problem hard to spot: gemini-2.5-flash answers with numbers rounded to exactly the values Keyword Planner uses (135,000, 14,800, 6,600, 5,400, 4,400, 3,600). The format imitates the real data, and an invented estimate dressed up as real data passes a visual check. None of its eight numbers was right, but every one of them looked like it could be.
On the UK test every estimate was at or above the real figure; when we ran the same test on Italian keywords, the errors went both ways. The size of the error ranges from 1.2 to almost 14 times, so there is no correction factor to apply. A general-purpose LLM is not a keyword research tool, and no newer version changes the nature of the problem: the volumes are not in the model, they are in a database.
Further reading: the method that starts from the data you already have is in keyword research: how to do it in 2026.
Where AI inside the tools really holds up
The model works when the data comes from somewhere else and the AI is left with the transformation work. These are tasks where the output can be checked in a few seconds and a mistake costs little.
- Grouping a list of a thousand keywords by intent, with volumes taken from the tool and not from the model.
- Writing titles and meta descriptions in bulk on an export, then correcting them by hand.
- Generating JSON-LD structured data from an existing page.
- Suggesting internal links based on pages already indexed, as Link Whisper does on WordPress.
- Turning a crawl into a priority list, reading columns a person would go through one at a time.
- Adapting a text from one market to another, which is different from translating it.
The UK picture helps show how much of this is already in use. According to the ONS Business Insights and Conditions Survey, self-reported use of AI among UK businesses with ten or more employees has risen from around 12% in late 2023 to around 35% in June 2026. The gap by size is still wide: 49% among businesses with 250 or more employees, 28% among those with fewer than ten. Improving business operations is the most common use, reported by over 60% of larger businesses, and adoption ranges from almost three in five businesses in information and communication to 13% in construction (ONS, Artificial intelligence in UK businesses: 2023 to 2026).
Further reading: which tasks are worth automating and which are not is in SEO automation.
What the AI layer costs: verified prices
Prices rarely appear in comparisons, and when they do they are out of date. These were read on the official pages on 24 September 2026. Unless stated otherwise, they are monthly prices billed annually, in US dollars, excluding VAT.
| Tool | Plan | Price | What it gives on the AI side |
|---|---|---|---|
| Semrush | SEO | $117.33/month (list price $139) | Basic AI monitoring, 500 tracked keywords |
| Semrush | Starter (SEO + AI Search) | $165.17/month (list price $199) | 50 prompts a day, 1 AI domain, MCP access |
| Semrush | Pro+ (SEO + AI Search) | $248.17/month (list price $299) | 100 prompts a day, cannibalisation analysis |
| Ahrefs | Lite | $129/month | No AI visibility features included |
| Ahrefs | Brand Radar (separate) | from $199/month | Organic prompts from the Ahrefs database, prompt packs from $50 to $250 |
| Surfer SEO | Discovery | $49/month | 120 documents, AI optimisation guidelines |
| Surfer SEO | AI Search Analytics | $82/month | 50 prompts updated daily, 5 Brand Spaces |
| Otterly.AI | Lite | $29/month | 15 prompts on ChatGPT, AI Overviews, Perplexity, Copilot |
| Otterly.AI | Standard | $189/month ($160 with annual discount) | 100 prompts on the same four engines |
One thing stands out when you scan the price column: the AI layer is almost never included. At Semrush it is a higher plan, and the difference between SEO and Starter is $47.84 a month, about $574 a year. At Ahrefs it is a separate subscription, so having search data and AI visibility together starts at 129 plus 199, that is $328 a month. Surfer has a plan dedicated to AI monitoring alone at $82.
The unit of measurement has changed, and nobody says so
For twenty years tools were sold by tracked keywords: the Semrush SEO plan gives 500, Ahrefs Lite 750, Advanced 5,000. AI visibility plans are sold by prompt: 15 on Otterly Lite, 100 on Otterly Standard, 50 on Surfer AI Search Analytics, 50 a day on Semrush Starter.
The jump in scale is two orders of magnitude. Anyone tracking three thousand keywords who moves to AI monitoring ends up choosing eighty questions, not importing a list. It is a selection job closer to qualitative research than to rank tracking, and it has to be done by someone who knows how customers talk.
There is also a problem with reading the data. Querying Google UK on 26 September 2026 for "ai tools", "ai seo tools" and "best seo tools", the AI Overview was present on all three queries. This time all three could be read on the first call, but that is not guaranteed: on the Italian versions of the same queries two of the three could not, and in our wider UK test two AI Overviews out of ten arrived asynchronously, so whoever queries them has to come back and ask again. A tool that does not handle that second call reports an absence that is not there.
Further reading: which tools really read AI Overviews in AI Overview tools, and the comparison by category in AI visibility tools.
The bill for a UK business
Every tool in this list is priced in dollars, so what a UK business pays depends on the exchange rate and on VAT. At the rate we use across this site (£1 = $1.3242), the Semrush Starter plan at $165.17 a month billed annually comes to about $1,982 a year, roughly £1,497, before VAT.
If you also publish in other languages, there is a point English-language comparisons never make: these models write better in English than in Italian, the other language we work in, and the gap shows in writing tasks more than in calculation. A draft generated in Italian by a tool built on English prompts arrives with syntax that needs rewriting, not correcting. The time saved on the first draft comes back in editing.
What no AI tool did for us
On a manufacturing project, Macropix, a commercial keyword moved from position 88 to position 2, with a 25.55% share of voice, ahead of Amazon. The work was an architecture review, a rewrite of the category pages by intent and a series of technical fixes. None of those decisions came out of a tool, AI or otherwise: the tools measured the before and after.
Further reading: the method and the full numbers are in SEO for manufacturing, along with our other work in projects.
What Google says, word for word
Google Search Central documentation is explicit about where the line is: using generative AI tools "to generate many pages without adding value for users may violate Google's spam policy on scaled content abuse". The test Google applies is little to no effort, little to no originality and little to no added value (Google Search Central, Guidance on using generative AI content). The same document says where generative AI is useful: researching a topic and adding structure to original content.
The distinction is not between written by a person and written by a machine; it is between pages that add something and pages that do not.
Further reading: data on indexing, rankings and clicks for AI content is in AI content and SEO.
Three situations and what each one needs
A business that publishes little and wants to know whether it exists in AI answers. Otterly.AI Lite at $29 a month with 15 prompts, plus Google Search Console, which is free. For under $400 a year you get a time series to reason on. Before buying, write down the fifteen questions a customer would ask an assistant: if you cannot think of fifteen, the problem is not the tool.
Teams producing content at volume. Surfer Discovery at $49 a month for briefs and optimisation, plus a general-purpose LLM used by hand for drafts. Keyword research stays out of the LLM, for the reason shown in the table at the start.
Anyone running SEO across several projects or for clients. A suite for search data and backlinks (Semrush SEO at $117.33 or Ahrefs Lite at $129 a month) plus a prompt tracker such as Otterly.AI Standard at $160 with the annual discount. The bill sits between roughly $250 and $400 a month (about £189 to £302), and at that point the biggest cost is still people's time.
For a selection made on your own case, together with optimisation for generative engines, there is Visilay's AI SEO service. Background on what has changed and what has not is in SEO and AI, while the strategy side is in generative engine optimisation and answer engine optimisation.
A note worth more than the price lists, for anyone setting the budget: the market has moved the unit of sale from the keyword to the prompt, and a prompt has no search volume. Nobody can tell you in advance what a citation inside an answer is worth, because the denominator does not exist. Whoever sells AI monitoring on subscription is selling a measure without a denominator, and the only honest way to use it today is as an internal time series, not as an industry benchmark.
AI SEO tools: frequently asked questions
No, and the line is clear: tools measure and speed things up, decisions stay outside them. In the Macropix project a commercial keyword moved from position 88 to position 2 thanks to an architecture review and a rewrite of the category pages by intent. The tools recorded the before and after; they did not choose what to change.
Not for volumes. On 26 September 2026 we asked gpt-4.1-mini and gemini-2.5-flash to estimate the UK search volume of eight industry keywords: two estimates out of sixteen fell within 30% of the Google Ads figure. On the keyword ai tools, which has 9,900 monthly searches, one model said 135,000. For grouping keywords by intent or generating variants to verify, on the other hand, it works.
The first uses a language model to do an existing SEO task faster, as Semrush, Ahrefs, Surfer SEO and Frase do. The second queries the assistants at regular intervals and counts how often a brand is cited: Otterly.AI, Profound, Peec AI, Rankscale, Ahrefs Brand Radar. The second does not optimise anything, it measures.
Google Search Console is still free and is the most reliable cross-check on what brings clicks. Ahrefs offers a limited Free plan and a Starter plan at 29 dollars a month, Surfer SEO starts at 49 dollars, Otterly.AI at 29 dollars with 15 prompts. Monitoring of AI answers is not included in any free plan: it is the feature vendors are building their new price lists on.
Google Search Central documentation says that generating many pages without adding value for users may violate the spam policy on scaled content abuse, and names the test it applies: little to no effort, little to no originality, little to no added value. The same document considers generative AI useful for researching a topic and adding structure to original content.
Entry plans deliberately give you few: 15 on Otterly.AI Lite, 50 on Surfer AI Search Analytics, 50 a day on Semrush Starter. Fifteen questions written the way a customer would ask them are worth more than a hundred generated automatically, because the time series only makes sense if the questions do not change over time.
