Does AI SEO Actually Work?
Partly. Google says optimising for AI search is still SEO, so the basics of clear pages, crawl access and trust do work. Evidence for AI-only tactics is thin, mostly observational and often from vendors. In the best public study we found (early 2025), AI assistants sent under 1% of visitors on average. Nobody can promise results, so test it on your own site for 90 days.
- Google says optimising for generative AI search is still SEO, and special AI files or markup are not needed.
- Most AI-specific studies are correlations or small experiments. Several come from firms that sell AI-visibility tools.
- AI assistants sent 0.17% of visitors in one 3,000-site study published February 2025, likely an undercount. Your own analytics may differ.
- Chalk Labs' own cases are search results measured in months. Third-party fast-lead claims are anecdotes, not benchmarks.
- Run a 90-day test with a fixed prompt list and a pass or fail rule before you commit a large budget.
So does AI SEO actually work?
It works to the extent that it is good SEO, and the proof for anything beyond that is weak. Google says that optimising for generative AI search is optimising for the search experience, and so is still SEO. 1 It also says you do not need new machine-readable files, AI text files or markup to appear in its AI features. 1
That is good news and a warning. Good, because pages that are crawlable, clear and trusted still matter. A warning, because anyone selling a secret AI-only trick has to show evidence, and public evidence for such tricks is mostly correlation. The table below grades the main claims.
| Claim | What the evidence shows | Strength |
|---|---|---|
| Good SEO basics still matter | Google states AI search optimisation is still SEO 1 | Strong: the engine's own documentation |
| You need to rank top 10 to be cited | Only 37.9% (37.1% counting organic links only) of URLs cited in AI Overviews also sat in the top 10, down from about 76% in an earlier study with a different method 11 | Moderate: large sample, one vendor, method changed |
| Brand mentions drive AI visibility | Branded web mentions correlate 0.656 to 0.709 with AI visibility across 75,000 established brands (DR above 40) 3 | Weak for cause: the authors say correlation is not causation |
| Publishing your own pages gets you named | For a new brand, 82% of new mentions came from answers citing the author's pages. For an established brand it was 6% 4 | Weak: 34 pages, one author, not controlled |
| llms.txt files help citations | Google says it neither helps nor harms visibility in Google Search 1 | Not supported for Google |
| AI traffic is large | 0.17% of visitors in a 3,000-site study from early 2025 5 | Moderate, but dated and likely an undercount |
What do you mean by work: visibility, traffic or leads?
Decide this first, because the answer differs for each. Visibility means an assistant names or cites your brand when asked. Traffic means people click through. Leads mean someone enquires or buys. A tactic can raise visibility and still produce no leads.
Most studies measure visibility, because it is easiest to count. Few follow a buyer from an answer to a sale. When a source says AI SEO works, ask which of the three it measured.
What does the visibility evidence say?
It says visibility is real but behaves differently from classic ranking. Ahrefs analysed 863,000 keyword results and 4 million AI Overview URLs and found 37.9% of cited URLs also appeared in the top 10. 11 An earlier Ahrefs study found about 76%, but the company says its method has since improved, so the two figures may not compare. 11
In a 75,000-brand study, Ahrefs found YouTube mentions had the strongest link to AI visibility, at about 0.737, and branded web mentions followed at 0.656 to 0.709. 3 The number of pages on a site correlated at only about 0.194. 3 The authors state that correlation is not causation, so this does not prove that raising mentions will raise visibility. 3 The brands were all established sites, so small businesses may behave differently.
One small experiment is useful because it tested a change. Ahrefs published 34 self-promotional pages and tracked 9,886 AI answers from February to May 2026. 4 For a new conference brand, 82% of new mentions came from answers that cited the author's pages. 4 But 43% of answers that cited a conference page still skipped the brand, so a citation did not always mean a mention. 4 The author says it is not proof, and the two brands are not a controlled comparison. 4
Does AI search send meaningful traffic yet?
Not on the best public numbers, though they are dated. A study of 3,000 sites published in February 2025 found AI chatbots accounted for 0.12% of views and 0.17% of visitors. 5 The top sites reached about 6%, and one reached 18%. 5 Ahrefs notes that many AI visits are recorded as direct traffic, so true numbers are likely higher. 5
Click behaviour on Google points the same way. In a Pew Research Center study of 900 US adults in March 2025, 18% of Google searches produced an AI summary. 6 People clicked a regular result in 8% of visits with a summary and 15% of visits without one. 6 Only 1% of visits with a summary clicked a link inside it. 6
So AI SEO may protect or win visibility without moving your traffic chart. Judge it by enquiries and by what buyers say, not by sessions alone. Our page on how much traffic ChatGPT sends goes deeper.
Do AI citations last?
Often not, which is why one-off work disappoints. Profound studied 883,000 pages across seven answer engines from September 2025 to September 2026. 7 It found the median page lost half its citation share 11 days after peaking. 7 About 22% of pages held on longer, with half-lives of 27 to 29 days. 7
Profound sells visibility tools, so use the figures as a rough guide. The lesson matches the Ahrefs experiment, where in roughly a quarter of cases a page was cited once for one query on one assistant and never again. 4 AI visibility needs upkeep, and a single launch is not a plan.
What do Chalk Labs' own cases show?
They show that search-led work can pay off over months. Over 16 months, Chalk Labs reports, the Meru Trophy shop in Bengaluru earned 27.3k Google clicks and 2.3M impressions at an average position of 6.9. 8 Its personal-brand site shilikajain.com reached number one for "fractional PR manager crypto" within a month of launch. 8
These are Chalk Labs' own results, reported by Chalk Labs, and they are search metrics, not revenue. They do not prove that AI assistants cite those sites more, and two cases cannot set a benchmark. They also cannot show that the work would repeat in a different field.
What about claims of leads in two weeks?
Treat them as anecdotes. Gushwork, a lead-generation firm, publishes case studies about its own clients, not Chalk Labs clients. One says Source Equipment got a first lead in 13 days, though the same page's metadata says 68 days. 9 Another says Respect Manufacturing got one in 12 days, while its body text says a qualified conversation took 35 days. 10
We do not know how many clients waited longer, or what the lead counts mean. The full breakdown is in can AI SEO produce leads in 15 days. Our own cases took months, and we found no verified source that gives a reliable time to first lead.
How do you test whether it works for you?
Run a 90-day test with rules set before you start. Pick 30 to 50 buyer questions, run them monthly in the assistants your buyers use, and record who is named and cited. Publish a small set of answer pages for those questions. Fix crawl access. Then compare month three with the baseline. Run each tracked question 3 times at baseline and at day 90, and count a question as 'named' only if you appear in at least 2 of the 3 runs.
Agree the pass mark in advance. For example: you are cited or named on more of the tracked questions, the cited answers are accurate, and at least some enquiries mention AI or a page you wrote. Treat small swings of a few questions as noise. Add a "how did you hear about us?" field to forms. Use the AI prompt list builder or read how to test whether your changes worked.
- Baseline (day 0): 40 tracked questions. We are named in X of 40.
- Pass if at day 90: named in at least X + 8 questions.
- Pass if: no tracked answer states a wrong price, location or claim about us.
- Pass if: enquiry form shows at least 1 AI-assistant source, or sales hears it on calls.
- Inconclusive if: visibility is up but enquiries are flat. Extend 60 days.
- Fail if: no change in naming or citation after 90 days and the pages are indexed.
What can no one promise?
No one can promise a ranking or a citation. Google says no one can guarantee a number-one ranking, and that if an adviser promises first place you should find someone else. 2 The same logic applies to AI answers, which change from one run to the next.
Chalk Labs prices three plans: Strategic Consulting at $1,000 one-time, Strategic Execution at $3,000 a month and Deep Execution at $4,000 a month. We do not guarantee rankings, citations or lead counts. We have a commercial interest in this topic, so check our claims against the sources above. The questions Google suggests asking any SEO are a fair test: what results to expect, in what timeframe, and how success is measured. 2
What published industrial results look like
Not Chalk Labs clients
These results were published by Gushwork for its own clients. Published results from Gushwork's own clients. Not Chalk Labs clients. We show them because they are the best public evidence that search-led inbound works for industrial companies, not as our work.
| Company | Industry | Published result | First lead |
|---|---|---|---|
| Source Equipment | Material handling equipment | 40+ qualified leads per month | 13 days |
| John Maye Company | Packaging equipment | 25+ leads in the first 30 days | Not stated |
| Respect Manufacturing | Contract manufacturing | 10+ qualified leads per month | 12 days |
| Midwest Power Products | Gearbox & motor distribution | 12 leads in the first month | Not stated |
| Cutting Edge Plasma | CNC equipment | 50+ leads generated | Not stated |
| Meridian Packaging | Industrial packaging distribution | 4+ qualified inbound leads per month | 75 days |
Frequently asked questions
Is AI SEO different from regular SEO?
Mostly it is regular SEO with extra attention to how assistants read and quote pages. Google says optimising for generative AI search is still SEO. The added work is answer-first writing, clear facts, and tracking whether assistants name you, which classic rank tracking does not show.
Does llms.txt work?
Not for Google. Google says such files neither help nor harm visibility in its Search. Other services may read them, but we found no verified evidence that they raise citations. Creating one costs little, so treat it as optional housekeeping and not a growth tactic.
Can anyone promise you AI citations?
No. Assistants change their answers between runs and between versions, and no one controls them. Google says no one can guarantee a number-one ranking. Be wary of any adviser who promises a citation, a ranking or a lead count, and ask for the method and the timeframe instead.
How do I know if AI SEO is working for my business?
Track a fixed list of buyer questions each month, record whether you are named and cited, and check that the answers are accurate. Then compare enquiries by source, including a self-reported field. If visibility rises but enquiries do not, extend the test before drawing a conclusion.
Is AI SEO worth it if AI sends little traffic?
It can be, if the same work improves your normal search results and how buyers see you in answers. Do not judge it by sessions alone. If your buyers rarely use assistants, the case is weaker, so check your own analytics and ask recent customers how they found you.
Sources
- Optimizing for generative AI search · Google Search CentralOpened October 20, 2026. Supports: Optimising for generative AI search is still SEO; no new machine-readable files, AI text files, markup or Markdown needed; llms.txt neither harms nor helps visibility in Google Search. Page last updated 2026-07-10.
- Do you need an SEO? · Google Search CentralOpened October 20, 2026. Supports: No one can guarantee a number-one ranking; find someone else if promised first place; questions to ask an SEO about expected results, timeframe and measurement. Page last updated 2026-06-05.
- AI brand visibility correlations · AhrefsOpened October 20, 2026. Supports: 75,000 brands (DR above 40); YouTube mentions about 0.737, branded web mentions 0.656 to 0.709, number of site pages about 0.194; authors state correlation is not causation. Published 12 December 2025.
- Self-promotional content AI SEO experiment · AhrefsOpened October 20, 2026. Supports: 34 pages, 9,886 AI answers, 7 February to 31 May 2026; 82% of a new brand's new mentions came from answers citing the author's pages versus 6% for an established brand; 43% of answers citing a conference page skipped the brand; in roughly a quarter of cases a page was cited once for one query on one assistant and never again; author says it is not proof and the comparison is not controlled.
- AI traffic study · AhrefsOpened October 20, 2026. Supports: 3,000 sites, published 6 February 2025: AI was 0.12% of views and 0.17% of visitors; top sites about 6%, one site 18%; many AI visits recorded as direct so true numbers likely higher.
- Google users are less likely to click on links when an AI summary appears in the results · Pew Research CenterOpened October 20, 2026. Supports: 900 US adults, March 2025: 18% of searches produced an AI summary; regular result clicked in 8% of visits with a summary versus 15% without; 1% clicked a link inside the summary.
- The half-life of an AI citation is 11 days · ProfoundOpened October 20, 2026. Supports: 883,000 pages across seven answer engines, September 2025 to September 2026: median half-life 11 days; about 22% of pages sustained with 27 to 29 day half-lives. Published 30 September 2026 by an AI-visibility vendor.
- Case studies · Chalk LabsOpened October 20, 2026. Supports: Chalk Labs' own reported results: Meru Trophy 27.3k Google clicks, 2.3M impressions, average position 6.9 over 16 months; shilikajain.com number one for one niche phrase within a month. Search metrics, not revenue.
- Source Equipment case study · GushworkOpened October 20, 2026. Supports: Gushwork-published claim of a first lead in 13 days; the page's metadata says 68 days. Third-party evidence for Gushwork's own client, not a Chalk Labs client.
- Respect Manufacturing case study · GushworkOpened October 20, 2026. Supports: Gushwork-published hero stat of 12 days to first lead; body text says 35 days to a qualified conversation. Third-party evidence for Gushwork's own client, not a Chalk Labs client.
- Ahrefs Brand Radar AI Overview citations and the top 10 · AhrefsOpened October 20, 2026. Supports: 863K keyword SERPs and 4M AI Overview URLs: 37.9% of cited URLs also in the top 10; earlier figure about 76% (July 2025) with an older method. Published 2 March 2026.