Last Updated on November 5, 2026
What this roadmap does
This 30-day roadmap helps a small business become visible, and accurately described, in answers from AI tools such as ChatGPT, Gemini and Perplexity. It treats AI SEO as an addition to normal SEO, not a replacement, and turns each tactic into a measurable test.
You will describe your business, test real customer questions across AI tools, fix gaps with clear answer-first content, earn genuine third-party mentions, and re-test after 30 days. It suits beginners who want a practical, evidence-led approach rather than unproven shortcuts.
Other names for the same job. GEO (generative engine optimisation) and AEO (answer engine optimisation) both mean optimising for AI answers. This guide mostly says “AI SEO”.
What counts as success. Not just how often your name appears. Measure whether AI visibility improves brand recognition, qualified enquiries, referrals and how accurately customers understand your business.
Be wary of guarantees. Nobody outside the AI companies can see exactly how an answer is assembled, so treat any promise of guaranteed results with suspicion. Every action below is labelled by how well supported it is.
If AI tools cannot access or interpret your pages, fix the basics first: make your site easy for AI tools to read.
The 30 days at a glance
Plan the 30 days in five phases, and fix your actions and owners on Day 10 before building anything.
| Phase | Days | What happens |
|---|---|---|
| Foundations | 1-2 | Describe the business, then check that AI tools can read your site |
| Baseline | 3-7 | Run 10 questions in 3 tools on 3 different days; log every run |
| Diagnose | 8-10 | Find gaps and their causes, rank actions, assign owners |
| Build | 11-24 | Two 7-day blocks: slow jobs first, then no page changes after Day 24 |
| Re-test | 25-30 | Repeat the same test on 3 different days; compare, then decide |
The baseline (Days 3-7) and the re-test (Days 25-30) use the same questions and conditions, so you can compare them like for like. Round one mainly tests your own pages; third-party mentions need later rounds.
Step 1 (Day 1): Describe your business
On Day 1, write 3-4 sentences that say what you offer, for whom, and the key facts: name, prices and location. Use the same facts on your website, profiles and listings, and adapt only the wording to each platform. If your facts differ from place to place, anything that reads them, human or machine, has to guess which is right.
Answer these questions first:
- What category does your business belong to?
- Which products or services set it apart?
- Who are your customers?
- Where do you operate?
- What evidence supports your expertise?
- Which other organisations, products or people are associated with your business?
Example: “Maple Street Bikes is an independent bicycle repair shop in Toronto, Canada, serving commuters and families. We service road, hybrid and e-bikes from $25 with same-week turnaround. Founded in 2015 by two certified mechanics, we are open Monday to Saturday.”
Save your text. You will reuse it in Step 5 when you update profiles, and in Step 3 to check whether AI tools describe you correctly.
See more: What Does Your Website Actually Know About Your Business? and 5 Questions to Ask When Improving Your Online Presence.
Step 2 (Day 2): Check that AI tools can read your site
Before you test anything, confirm that your key pages can be found, read and shown. If they are blocked, your baseline in Step 3 will say nothing useful. Google’s own guidance gives the basic rule for its AI features: a page must be indexed and eligible to appear in Google Search with a snippet, with no extra technical requirements.
Work through this list for your 5-10 most important pages. Ask your web person for help with any item you cannot check yourself.
- Indexed. In Google Search Console (free), inspect each page with the URL Inspection tool and confirm Google has indexed it.
- Not blocked.
robots.txtis a small text file atyourdomain.com/robots.txtthat tells crawlers which pages they may visit. Google lists allowing crawling in robots.txt, and in any content delivery network (CDN) or hosting set-up, as a basic best practice. - Facts in text. Prices, services, opening hours and location appear as normal text on the page, not only inside images or scripts.
- Readable without tricks. If a page looks empty when JavaScript is off, ask your web person to check it. Google can process JavaScript content, but JavaScript sites need extra care.
- Included in Google’s AI features. Google says a site must be included in Search generative AI features in Search Console to be eligible for display there. Check that setting.
- Crawlers chosen on purpose. Some crawlers fetch pages so a tool can show or cite them in answers. Others collect data for training. They are separate decisions. Google, for example, uses Googlebot for Search and Google-Extended to limit AI training and grounding in some of its other systems. Each AI company documents its own crawlers, so ask your web person which ones your
robots.txtallows and blocks.
Fix what you can now. If something cannot be fixed this week, write it down and note it next to your results in Step 3, because it limits what you can conclude.
Step 3 (Days 3-7): Measure where you stand
Write 10 customer questions, ask each one in ChatGPT, Gemini and Perplexity on three different days, and log every run the same way. Answers vary from run to run, so a single run can mislead you.
1. Write your 10 questions. Mix two kinds, roughly half and half:
- Branded questions name your business: “Is Maple Street Bikes any good?” They show whether AI describes you correctly.
- Unbranded questions do not: “Where can I get an e-bike serviced in Leeds?” They show whether AI considers you at all.
Use real customer wording from emails, reviews and enquiries. Set aside generic questions with no business purpose, because they produce zeros that are not real problems.
2. Run them on Days 3, 5 and 7. That is 10 questions × 3 tools × 3 days = 90 runs. If that is too much for round one, use 5 questions or 2 tools, but keep exactly the same set for the re-test in Step 7.
3. Log every run in a spreadsheet, one row per run. Record the conditions, because they change the answer and you must repeat them at the re-test:
- date, and the tool with its mode (for example, web search on or off)
- logged in or logged out
- location and language
- the exact question
Then record these seven results:
| Measure | What you check | What it tells you |
|---|---|---|
| Mention | Did your brand name appear? | AI recognises or discusses your brand |
| Recommendation | Was your business suggested as an option for the need? | AI considers you relevant to the question |
| Citation | Was your own site named or linked as a source? | AI uses your pages as evidence |
| Link | Was there a clickable link to your site? | People can visit you directly |
| Other sources | Which other sites were cited? | Where AI gets its information about your field |
| Accuracy | Do the facts match your Step 1 description? | AI describes your business correctly |
| Sentiment | Was the tone positive, neutral or negative? | How AI frames you |
Treat the result as a baseline, not a verdict. Don’t set numeric goals in round one; build a picture over 2-3 rounds first.
Before running tests, make sure you understand the fundamentals of AI SEO: AI search is still built on visibility, trust and clear structure.
Step 4 (Days 8-9): Find out why you’re missing or misdescribed
For every question where you are absent or described wrongly, write down who is named instead, which sites are cited, and one likely reason. That reason is your hypothesis: a best guess you will test, for example “competitors are cited via review sites and we have no reviews there”.
Compare yourself with the competitors AI names. Open their pages next to yours and ask:
- Do they provide independent reviews?
- Do they compare several providers?
- Do they explain product features?
- Do they contain original research?
- Do they show first-hand experience?
Find the real wording of customer questions. Look at Google’s “People also ask” boxes (the related questions shown in search results), customer emails and reviews, and forums in your field. Rewrite your questions from Step 3 if real customers phrase them differently.
When AI gets your facts wrong, fix the source, not just the symptom:
- Find out where the wrong fact comes from. Tools that show sources, such as Perplexity, usually reveal the page they used.
- If the source is your own site or an old profile, correct it. If it is a third-party page, such as a directory or an old article, ask its owner to update it.
- State the correct fact as plain text on your site (price, opening hours, location) and keep your Step 1 description identical everywhere.
- Report serious errors through the tool’s feedback option, where it has one.
- Re-test after a few weeks. AI answers do not update at once: Google notes that recrawling can take anywhere from several days to several months, and there is no guaranteed way to force a correction.
End Day 9 with a short list of gaps, each with its hypothesis.
Step 5 (Day 10): Choose and rank your actions
Pick actions from the tables below, score each for impact and effort, and start with the best ratio. Every action carries an evidence label so you know how much to trust it.
How strong is the evidence? Google’s guidance for AI Overviews and AI Mode says the same SEO best practices apply and no special optimisation is needed. Its guide to generative AI features says unique, non-commodity content probably matters most in the long run, and it lists tactics to ignore. That covers Google only. This guide does not cover official guidance from other AI companies, so treat effects on ChatGPT, Perplexity and others as plausible but unproven, and check them with your own tests.
Do first
| Action | Evidence | Why |
|---|---|---|
| Create unique, first-hand content | Supported by Google’s guidance | Google contrasts first-hand experience with content that recycles what others already said |
| Show real experience: who wrote the page and what they have done | Supported by Google’s guidance | It makes the first-hand point visible to readers |
| Keep key facts (prices, features, hours) as plain text | Supported by Google’s guidance | Google lists important content in textual form as a best practice |
| Keep listings up to date and consistent: Google Business Profile, Merchant Center, directories | Supported by Google’s guidance | Google says these help local and product details appear in AI responses |
| Make pages easy to find through internal links with descriptive link text (“learn the basics of a balanced diet”, not “click here”) | Internal links supported by Google’s guidance | Readers and crawlers both follow them |
| Use plain question-style headings, answer first, one idea per section | Reasonable, untested for AI tools | Google endorses clear headings for readers; answer-first is not a Google requirement |
| Earn genuine third-party mentions: reviews, news, directories | Plausible, and authenticity matters | Google says AI features can reflect what is said about products across the web, and that seeking inauthentic mentions is not as helpful as it might seem |
| Name and link the source of facts (“the 2025 X study”, not “most people prefer X”) | Reasonable, untested | Readers and tools can check the claim |
| Show dates and refresh old pages | Reasonable, untested | Out-of-date prices and hours are easy to repeat wrongly |
One warning on question-style headings: write one section per real customer question, but do not create a separate page for every variation of a question. Google says doing that mainly to manipulate rankings or AI responses breaks its scaled content abuse policy.
Only if you have spare time. None of these is needed to appear in Google’s AI features, so do them after the actions above.
| Extra | Evidence | Note |
|---|---|---|
| Schema markup (structured data) | Not required by Google for AI features | There is no special schema.org markup for AI features. It can still help with rich results (enhanced listings in Google Search) and must match the visible text |
| llms.txt and other AI text files | Google Search ignores them | Google says creating them for other services is fine, but their benefit elsewhere is unproven |
| Splitting content into tiny “chunks” | Google says it is not required | Write clear sections for readers, at whatever length suits the topic |
| FAQ blocks | Unproven | Add one only if real customers ask those questions, not for AI |
Never: buy or fake mentions, reviews or endorsements.
Cover three kinds of content. Explanatory pages help people understand a topic, comparative pages help them evaluate alternatives, and evidential pages demonstrate your claims, expertise and experience. A web hosting company, for example, should not only explain what hosting is. It should also compare hosting options and publish evidence of its own experience.
Score and rank. Give each action a score from 1 to 5 for impact (how much it could help) and for effort (time and cost; 5 is hardest). Divide impact by effort and start with the highest result. If two actions tie, start the one that pays off most slowly, which is usually third-party mentions.
| Example action (invented) | Impact | Effort | Impact ÷ effort |
|---|---|---|---|
| Rewrite the pricing page with answer-first sections | 4 | 2 | 2.0 |
| Ask five customers for independent reviews | 4 | 3 | 1.3 |
| Add FAQ schema | 1 | 2 | 0.5 |
Assign owners. Give every action you choose one owner, one measure of success and one re-test date between Day 26 and Day 30.
Step 6 (Days 11-24): Build in two 7-day blocks
Spend Days 11-24 working through your ranked actions in two 7-day blocks, and start the slow jobs on Day 11.
Block A (Days 11-17): slow jobs first
- Start outreach for genuine reviews, mentions and directory listings. These take the longest to show up.
- Claim, correct and align your profiles and listings with your Step 1 description.
- Rewrite your highest-ranked page. The quick start below shows how.
Block B (Days 18-24): finish and freeze
- Finish the remaining rewrites, internal links, named sources and dates.
- Request a recrawl of changed pages in Search Console, since Google may take days to months to recrawl a page.
- Stop changing pages on Day 24, so the re-test measures a stable site.
Throughout both blocks, keep a change log with one row per change: date, page, what changed and which action it belongs to. If you can, leave a few similar pages unchanged as a comparison group.
Quick start: one page in 4 days
- Day 1: Split one important page into sections, each with a plain, question-style heading.
- Day 2: Add an answer-first opening to each section.
- Day 3: Read each section on its own. Could someone understand it without the rest of the page? Add any missing context.
- Day 4: Add internal links with descriptive link text. Schema markup is optional and comes last (see Step 5).
Step 7 (Days 25-30): Re-test and decide
Repeat your baseline on Days 26, 28 and 30 exactly as before, then compare the two rounds and decide what to do with each action.
Re-test. Use the same questions, tools, modes, logged-in state, location and language as in Step 3, on three different days. Fill in the same seven measures. If you left some pages unchanged, compare them with the changed ones.
Decide, action by action:
- Keep it if its success measure moved the right way.
- Expand it if it worked on one page and similar pages exist.
- Revise it if nothing moved but your hypothesis from Step 4 is still plausible.
- Stop it if nothing moved and you cannot say why it should, or the cost is too high.
Read round one with care. Changes to your own pages may not have reached AI answers yet, because recrawling can take days to months. Third-party mentions can take months. Round one therefore mostly tests your on-site changes, and a flat result is not a failure. Check accuracy as well as visibility: a mention that gets your facts wrong is a problem to fix, not a win.
Plan rounds two and three. Repeat the same test around Day 60 and Day 90. This is a suggested rhythm, not a rule. Between rounds, add a “How did you hear about us?” question to your enquiry form, so AI influence can show up in enquiries and referrals, not only in answers.
Free tools to track progress
Google’s Search Console has a free Generative AI performance report, but it measures Google’s AI features only. It does not replace your manual tests in ChatGPT, Gemini and Perplexity.
- What it shows: impressions in AI Overviews and AI Mode, grouped by page, country, device or date, with an export button. An impression means a link to your site was shown in an AI feature.
- What it leaves out: other AI tools, and any query or click data. Its help page lists neither, so you cannot see which questions triggered an appearance.
- Who can see it: Google’s help page says it was rolled out to all websites worldwide as of 31 August 2026. If you do not see it, your site may not have received enough impressions yet, or you may have excluded it from Google’s AI features.
- Where else it counts: AI feature traffic is also included in the Web search type of the normal Performance report.
- Free by-hand checks: whether you appear in featured snippets and in “People also ask”, and the “How did you hear about us?” answers from Step 7.
You do not need paid tools. They can automate the logging, but everything in this roadmap works by hand. Be wary of any tool that claims to use Google’s internal ranking data: Google says no third-party tool has access to its internal ranking or AI systems.
Good to know
Most mistakes in round one come from over-reading the numbers.
- Clicks are not the whole story. AI influence often shows up as brand searches, direct visits and enquiries rather than clicks.
- No numeric goals in round one. Build a baseline over 2-3 rounds first.
- One run is not a result. Answers differ by tool, mode, location, language and day, so always log the conditions.
- Zeros are not always problems. Generic questions with no business purpose can show zeros that mean nothing. Set them aside.
- Check accuracy, not just mentions. A mention with wrong facts needs fixing.
- Be patient. Third-party mentions take months, and recrawling your own changes can take days to months.
- Re-check the guidance. This field changes fast. Look at the latest versions of the Google pages listed under Sources before making big decisions.
Glossary of AI SEO terms
Sixteen terms used in this guide, in plain words.
| Term | Meaning |
|---|---|
| Baseline | Your starting measurements, taken before you change anything |
| Branded / unbranded question | A question that names your business, and one that does not |
| Chunking | Splitting content into very short, self-contained pieces. Google says this is not required |
| Citation | An AI answer naming or linking a website as a source. A “citation of your site” means it named or linked yours |
| Crawlable | A page that search engines and AI tools are allowed and able to access |
| GEO / AEO | Generative engine optimisation and answer engine optimisation: other names for optimising for AI answers |
| Hypothesis | Your best guess why something happens, which you then test |
| Impression | In Search Console, a link to your site shown to a user in a Google AI feature |
| Indexed | Stored in a search engine’s index, so the page can be shown in results |
| KPI (success measure) | A number that shows whether an action worked |
| llms.txt | A file meant to guide AI tools. Google Search ignores it, and its value elsewhere is unproven |
| Mention | Your brand name appearing in an AI answer |
| Prompt | A question you type into an AI tool |
| Re-test | Repeating the same measurements later, under the same conditions, to compare |
| robots.txt | A text file at yourdomain.com/robots.txt that tells crawlers which pages they may visit |
| Schema markup | Code, invisible to readers, that labels page content for machines. Also called structured data |
Sources and further reading
Sources
- AI features and your website, Google Search Central
- Optimizing your website for generative AI features on Google Search, Google Search Central
- Generative AI performance report (Search), Search Console Help
- Introducing Search Generative AI performance reports in Search Console, Google Search Central Blog
Further reading
- New to SEO? Start with five SEO basics that actually matter before building this roadmap.
- Make your site easy for AI tools to read
- The fundamentals of AI SEO
- Books that can help: SEO & AI – A Practical Beginner’s Guide and SEO All You Need to Know: Get Yourself and Your Website Found!