How to Optimise Your Website for AI Search: A 10-Step Checklist

To optimise your website for AI search, do three things in order:

• Let the AI crawlers reach your pages,
• Structure those pages so a model can lift a clean answer out of them,
• Build enough evidence elsewhere online that a model trusts naming you.

Most business owners I speak to are stuck on the first one without knowing it, because the bot that decides whether you appear in a ChatGPT answer is not the same bot most people think they are allowing.

Below are the ten steps in the order I run them, with the exact files and settings. If you would rather hand it over, that is my AI search optimisation services for UK businesses.

Key Takeaways

• AI search visibility sits on top of SEO, not instead of it. A slow, badly structured site does not become citable because you added schema.
• Most sites are blocking the wrong crawlers by accident. Training bots and search bots are different bots. Blocking GPTBot does not remove you from ChatGPT search, and blocking ClaudeBot does not remove you from Claude’s answers. You have to name the search agents separately.
• Answer capsules matter more than word count. A direct 40 to 60 word answer under every heading gives a model something clean to lift.
• Being mentioned off your own site is the step most people skip. When I ran this on myself, a Reddit profile with two followers still got picked up as a source.
• Realistically it takes months, not weeks. My own experiment took six months before I showed up by name in AI answers.
• One original number you can publish turns a generic guide into something an AI has a reason to cite.

In short: AI search optimisation is not a new discipline bolted onto SEO. It is the same foundations, plus structure that a model can parse, plus evidence of you existing somewhere other than your own website.

Before you start: two things to confirm before step one

You need two things before any of this works: access to Google Search Console for your domain, and the ability to edit your website, either directly or through someone who can. If you have neither, stop here and email whoever built your site. Ask for admin access to the CMS and to be added as a user in Search Console.

I am being blunt about this because I have watched business owners spend three months on content while their site quietly sat behind a setting nobody had checked since launch.

If your site was built by an agency you no longer work with, this is the moment to get your logins back. Ask for: CMS admin access, hosting or FTP access, domain registrar access, and Search Console ownership. That is a reasonable request and any decent developer will hand it over without a fuss.

The time expectation: about two hours to work through the first seven steps if you have access and a site with under 30 pages. Steps 8, 9 and 10 are ongoing rather than one afternoon.

If you do not control your own website at all, I have written a separate technical SEO checklist for business owners who do not control their own website that covers what to ask for and how to phrase it.

Step 1: Check that AI crawlers can reach your site, or nothing else here matters

Before you change anything, find out whether AI crawlers are visiting you at all. Open your server logs or your hosting analytics and search for these user agents: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot and Google-Extended. If none of them appear in the last 30 days, something is blocking them.

Here is the part that trips people up. Those bots do different jobs, and treating them as one group is why so many sites get this wrong.

There are three families:

• Training crawlers. GPTBot (OpenAI), ClaudeBot (Anthropic), Google-Extended (Google). These collect content to train or ground foundation models. They crawl in bulk and they respect robots.txt.
• Search and retrieval crawlers. OAI-SearchBot (ChatGPT search), Claude-SearchBot, PerplexityBot. These are the ones that decide whether you appear inside an AI answer. If you only allow one group, make it this one. OpenAI is explicit about this in its own overview of OpenAI crawlers: each setting is independent, so you can allow one and block another.
• User-triggered fetchers. ChatGPT-User and Claude-User. These fire when a person explicitly asks the assistant to go and look at a URL. Blocking these is almost never what you want, because it breaks something your own visitor asked for.

Blocking a training crawler and blocking a retrieval crawler are not the same decision, and they do not have the same consequence. You can refuse to feed model training while staying fully eligible for citations. Plenty of publishers make exactly that choice. What you should not do is block the retrieval crawlers by accident, which is what usually happens.

How to check without touching a log file. If server logs are beyond you, do this instead: open your site in a browser and go to yourdomain.co.uk/robots.txt. Read what is there. Then check whether your CDN or security plugin has its own bot rules, because Cloudflare, Sucuri and similar tools often have an AI bot toggle that overrides robots.txt entirely. A lot of sites built or last audited before 2023 are blocking these bots through a security setting nobody remembers switching on.

In short: your robots.txt can say yes while your CDN quietly says no. Check both, because the file is not the only gatekeeper.

Step 2: Fix your robots.txt so you stop blocking the crawlers you actually want

Open yourdomain.co.uk/robots.txt. Add explicit rules naming each AI crawler you want to allow, and always finish the file with a sitemap line. Then validate it before you go live, because a syntax error here can block Googlebot and take out your ordinary search traffic alongside everything else.

Here is a real before and after.

Before (this is the pattern I find most often, usually left behind from a staging site or added by a plugin):

User-agent: *
Disallow: /

Sitemap: https://yourdomain.co.uk/sitemap.xml

That first block tells every bot on the internet to stay out of every page. It is two lines and it undoes everything else in this checklist.

After, if you want full AI visibility:

# Search and retrieval crawlers: these decide if you appear in AI answers
User-agent: OAI-SearchBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

# User-triggered fetches: someone asked the assistant to read your page
User-agent: ChatGPT-User
Allow: /

User-agent: Claude-User
Allow: /

# Training crawlers: your call, allowing them is the simpler choice
User-agent: GPTBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: Google-Extended
Allow: /

# Keep admin areas out of it
User-agent: *
Disallow: /wp-admin/
Disallow: /checkout/

Sitemap: https://yourdomain.co.uk/sitemap.xml

Allow versus Disallow, in plain English. Disallow: / means “do not crawl anything on this site”. Allow: / means “you may crawl everything”. Disallow: /wp-admin/ means “everything except this folder”. The rules apply to whichever User-agent line sits directly above them, which is why each bot gets its own block.

If you want to opt out of training but stay citable, change Allow: / to Disallow: / under GPTBot, ClaudeBot and Google-Extended, and leave the search crawlers on Allow. Blocking Google-Extended does not affect Googlebot, your ordinary Google rankings or your indexing. It only opts you out of Gemini training and grounding, and Google says so directly in its list of common Google crawlers, which confirms that Google-Extended and GoogleOther do not affect search rankings.

If you want to understand exactly how the rules are interpreted rather than guessing, Google publishes a plain reference on how Google interprets the robots.txt specification, and the syntax is the same standard the AI crawlers follow.

Two honest caveats. First, robots.txt is a request, not a wall. Compliance is opt-in, and some crawlers have been documented ignoring it entirely. If a bot is genuinely hammering your server, that is a job for your firewall, not this file. Second, a user-agent string can be faked, so if you are making blocking decisions based on your logs, verify with a reverse DNS lookup rather than trusting the header.

Validate the file before you publish it. A misplaced slash can cost you your Google traffic, and that is a much more expensive mistake than not being in ChatGPT.

Step 3: Give every page one clear job so a model knows when to quote it

One page, one question, one intent. AI models pull a passage, not a website, so a page that tries to answer six things gives the model nothing specific enough to lift. If you cannot say in one sentence what question a page answers, it is doing too much.

The most common version of this problem is the bloated service page. Here is a real shape I see constantly on plumbing, clinic and trades sites:

Before: one page called “Our Services” covering emergency callouts, boiler installation, bathroom fitting, underfloor heating, landlord gas safety certificates and the areas covered, at 2,400 words, with a contact form at the bottom.

After: six pages. Emergency plumber in [town]. Boiler installation costs and options. Bathroom fitting. Underfloor heating. Landlord gas safety certificates. Areas covered. Each one 600 to 900 words, each answering one question properly, each internally linked to the others.

The second version wins for a reason that has nothing to do with keyword stuffing. When someone asks an AI “how much does a landlord gas safety certificate cost in Manchester”, the model wants a passage that answers exactly that. On the before version, the relevant content is 1,800 words into a page about six other things. On the after version, it is the whole page.

The quick audit. List every page on your site in a spreadsheet. Next to each one, write the single question it answers, in the customer’s words. Any page where you cannot do that, or where you write two questions, is a page to split.

Step 4: Write an answer capsule under every heading so AI can lift it cleanly

Directly under each H2, write a 40 to 60 word paragraph that answers that heading completely on its own. It has to make sense to someone who reads only that paragraph and nothing else on the page. This is the single highest-return formatting change you can make, and it takes an afternoon.

This is the AEO part, answer engine optimisation, and it is really just writing the conclusion first. Journalists have done it for a century. We got out of the habit because we were writing for a scroll, not for an extraction.

Before, from a plumber’s FAQ page:

How much does a boiler service cost? There are a lot of factors that go into boiler servicing and every home is different. We have been serving customers in the area for over 20 years and pride ourselves on transparent pricing. Get in touch and we will be happy to give you a quote tailored to your circumstances.

Nothing in there can be quoted. A model reading that has no answer to hand back.

After:

How much does a boiler service cost? A standard gas boiler service in the UK typically costs between £70 and £120, depending on your area and boiler type. We charge £85 for a full service including a flue gas analysis and a Gas Safe certificate. Landlords booking a CP12 at the same time pay £110 for both.

Same length. One is liftable, one is not.

A clinic version works the same way. Instead of “recovery times vary from patient to patient”, write “most patients return to desk work within 5 to 7 days and to the gym at 3 weeks, though your surgeon will confirm based on your procedure.”

You are not giving away the consultation. You are proving you know the answer, which is the entire basis on which a model decides who to name.

Step 5: Rewrite your headings as the questions your customers actually ask

Industry phrasing and customer phrasing are almost never the same, and models match on customer phrasing. Go through your headings and rewrite each one as the question a person would type or say out loud. You can do the whole site in the next 10 minutes.

Some real contrasts:

What the business writes	What the customer asks
Our Composite Bonding Treatment	How long does composite bonding last?
Comprehensive Pest Management	How do I get rid of mice in a terraced house?
Bespoke Wealth Planning Solutions	Do I need a financial adviser if I have £50,000 to invest?
Vehicle Diagnostics	Why is my engine management light on but the car drives fine?

The rewriting formula. Take your existing heading. Ask: what is the person’s actual situation, and what do they want to change about it? Then write it as they would say it. [Question word] + [their situation] + [the outcome they want].

Where do you find the real phrasing? Three free places. Your own Google Search Console query report, which is the most underused free tool most business owners own. The People Also Ask box on a Google search for your main service. And the questions people ask in Reddit and Quora threads about your industry.

That last one is not a throwaway suggestion. When I worked with a healthcare brand, I spent proper time reading forums to understand how patients actually described their concerns, then built content around that language rather than around clinical terminology. Domain authority went from 15 to 28 in six months, and the brand eventually started appearing in LLM-generated answers. The forum reading was not the boring prep bit. It was the whole strategy.

Step 6: Add the schema that actually helps a small business get picked up

Four schema types cover almost every small business: Organisation, LocalBusiness, FAQPage and Article. Add them as JSON-LD in your page’s head or through your SEO plugin, then test every page with a free validator before you consider it done.

Schema markup is structured data, which is a machine-readable summary of what is on the page. The vocabulary is public and free to read at schema.org, and the four types above have their own reference pages: Organization, LocalBusiness, FAQPage and Article. It does not change what a visitor sees. It tells a crawler “this is a business, this is its address, these are questions and these are their answers”, so nothing has to be inferred.

Here is a copy-paste FAQPage example. Swap the questions and answers for your own:

json

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "How much does a boiler service cost?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "A standard gas boiler service in the UK typically costs between £70 and £120. We charge £85 for a full service including a flue gas analysis and a Gas Safe certificate."
    }
  }, {
    "@type": "Question",
    "name": "How long does a boiler service take?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "About 30 to 45 minutes for a standard combi boiler with no faults found. Older or back-boiler systems can take up to an hour."
    }
  }]
}

Which type goes where. Organisation goes on your homepage and About page. LocalBusiness goes on your homepage and any location pages, and it wants your address, opening hours and phone number. FAQPage goes on any page with a genuine question and answer section. Article goes on blog posts, with the author and the published date filled in.

The warning. Do not mark up content that is not visible on the page. If your FAQPage schema contains three questions and the page only shows two, that is a mismatch, and at best the markup gets ignored. At worst it counts against you. The schema describes the page, it does not replace it.

Run every page through a schema validator before you move on. Google’s Rich Results Test and the Schema Markup Validator are both free, and both will tell you exactly which line is broken. Google also documents which structured data features it supports, which is worth a scan before you invest time in a type that does nothing for your sector.

Step 7: Make your business details identical everywhere so a model trusts who you are

Your business name, address, phone number and description need to match character for character across every platform you appear on. AI models cross-reference sources. When three listings disagree about your address, the model has no way to know which is right, so it tends to trust none of them.

The UK-specific places to check, in the order I work through them:

• Google Business Profile
• Bing Places
• Yell
• Trustpilot
• LinkedIn company page
• Companies House, which is worth checking because it is a high-trust public record, it is freely searchable by anyone including AI crawlers, and it is frequently out of date
• Your sector directories: Checkatrade, Rated People, the CQC register, the FCA register, Doctify, the Law Society, whichever applies to you

The details that quietly drift: an old phone number on a directory you forgot you signed up to, “Ltd” on one and “Limited” on another, a first-floor suite number that appears on some listings and not others, and a description that was written in 2019 and describes a business you no longer run.

I saw the value of this most clearly on a salon business I worked with. The budget was tiny, the equivalent of about £280 over six months, so there was no money for anything clever. Getting the Google Business Profile complete, consistent and actually maintained did more for that business than any campaign we ran. It now takes bookings through existing clients, referrals and the Business Profile without paying for ads. Consistency is unglamorous and it is free.

Write your name, address, phone and 50-word description in a single document. Copy from that document every time, forever. That is the whole system.

Step 8: Get mentioned where AI already looks, not just on your own site

If your website is the only place you exist online, you have given the model one source instead of ten. Models weigh multiple sources before naming anyone. The places worth your time in the UK are Reddit, Quora, Wikipedia if you genuinely qualify, trade press, and your sector’s directories.

This is the step people skip, and it is the one that changed everything when I tested this workflow on myself.

I ran the full thing on my own business rather than a client’s, in public, for six months, and I talked through the results on stage at SEOCharity. The test I used at the end was simple: ask a clean-slate AI, one with no memory of me and no history of me talking to it, a question my ideal client would ask. Something like “who works at the intersection of SEO, content and AI in the UK?”

I came back first on that list. Perplexity had me at number twenty out of more than a hundred sources, citing my site and an interview I had given. Google’s AI Mode described me as leading its selection of verified professionals, which I read three times before I believed it.

Here is the part I want you to take from this. My Reddit profile, one of the places I had been consistently answering real questions, had two followers. Two. And it still got picked up as a source.

That is the whole lesson. You are not chasing followers, reach or a viral post. You are building a situation where an AI model goes looking for evidence about you and finds the same consistent story in five different places. A two-follower profile that answers real questions honestly is worth more to this process than a thousand followers who never see you say anything useful.

What this looks like in practice for a business owner. Find the three subreddits and the handful of Quora topics where your customers ask questions. Answer properly. Not a link drop, not a pitch, an actual answer that would be useful even if nobody ever clicked through to you. Do that a few times a month. Contribute to your trade press when they ask for comment. Claim your directory listings.

Use your own experience here. First-person evidence from someone who has done the thing beats generic advice every single time, both for the human reading it and for the model deciding whether you are worth naming. If you have been fitting bathrooms for 15 years, you know things that no content writer knows. Say those things, in public, where they can be found.

If you want the longer version of how this fits together across platforms, I have written about search everywhere optimisation for local businesses, and the full four-phase version of my own experiment is in the GEO workflow I tested on myself.

Step 9: Add one number only you can publish so you become a source

Original data turns a generic guide into something an AI has a reason to cite, because it cannot get that number anywhere else. Most small businesses are sitting on publishable data and have never thought of it as data. You can usually produce one in an afternoon from records you already keep.

Three examples of what counts, all achievable without a research budget:

1. A client result with a real before and after. “We took a client’s domain authority from 15 to 28 in six months.” Specific, dated, verifiable in principle.
2. A before and after stat from your own operations. “Switching to a two-stage booking form cut our no-show rate from 22% to 9% across 400 appointments last year.”
3. A count from your own records. “Of the 312 boilers we serviced in 2025, 41% of faults were caused by low system pressure, and 12% were the same three thermostat models.”

That third type is the one people underestimate. A pest control firm knows which months mouse callouts spike. A dentist knows the average age of a first-time composite bonding patient. A recruitment consultant knows how long placements actually take in their sector. Nobody else has published those numbers, which is exactly why they get cited.

How to publish it properly. State the number, state the sample size, state the date range, and say how you counted it. “Based on 312 service jobs completed between January and December 2025” is what turns an assertion into a source. Without the methodology line, it reads as a claim, and models and humans both discount claims.

One rule I hold to: never publish a number you have not checked in your own records. An invented statistic will eventually be quoted back at you, and you will have to defend it.

Step 10: Put a credentialled human on the page so a model has a reason to trust it

Every substantive page needs a real named author, a short bio with actual credentials, and outbound links to the sources you relied on. Pages that cite well tend to get cited. This is E-E-A-T made practical rather than theoretical.

E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness. It is Google’s framework for quality raters, not a ranking factor you can switch on, but it describes the qualities that make a page look like a reliable source, and those are the same qualities a model is weighing. Google sets out what it actually means by it in its guidance on creating helpful, reliable, people-first content, which is worth twenty minutes of your time and is written for people, not developers.

The four things to add:

1. A real byline. A name, linking to an author page. Not “admin”, not “the team”, not nothing at all.
2. A short bio under the article. Two or three sentences with the credentials that matter in your field. Gas Safe registration number. GDC number. Years in the trade. Where you trained. Not a personality paragraph, a credentials paragraph.
3. Outbound citations. Link to the standard, the regulator, the study, the manufacturer’s guidance. Showing your working is a trust signal. It also stops you being the only unsourced page in a topic where every competitor cites theirs.
4. A published and updated date, visible on the page. Not just in the schema.

The version of this I find persuasive is that AI models are doing what a careful reader does: asking who wrote this and why should I believe them. If the page cannot answer that, it does not get used, however well structured the rest of it is.

Common mistakes I see business owners make with this

• Adding schema before fixing access. Perfectly marked-up pages that no crawler can reach are perfectly invisible.
• Blocking training bots and assuming that keeps you out of AI search. It does not, and the reverse is worse: blocking the search bots while allowing training means you feed the model and get no credit.
• Marking up content that is not on the page. It is the fastest way to have your schema ignored.
• Treating the ten steps as a one-off. Steps 1 to 7 are a project. Steps 8, 9 and 10 are a habit.
• Waiting for perfect before publishing anything off-site. Two followers, remember.
• Copying the same answer onto Reddit, Quora and your blog. Recognisably you, everywhere, is the goal. Copy-pasted is not the same thing, and forum communities will spot it immediately.

How long before something changes?

Expect months, not weeks. In my own experiment it took six months of consistent work across all ten steps before I started showing up by name in AI answers. Perplexity tends to reflect changes fastest, sometimes within weeks. ChatGPT is slower. Anyone promising results in a fortnight is not describing a real workflow.

What the timeline actually depends on:

• Your domain authority. A site with existing trust gets re-crawled more often and gets picked up faster.
• Your existing mentions. If you already appear in trade press and directories, you are starting from a higher base than someone with a website and nothing else.
• Crawl frequency. A site publishing monthly gets visited more than a site that has not changed since 2022.
• How competitive your topic is. “Emergency plumber in a small town” is a very different problem from “best CRM software”.

How to test whether it is working. Do what I did. Open a clean-slate AI session, one with no memory of you and no history of you talking to it, and ask the question your ideal customer would ask. Not “tell me about [your business]”, which proves nothing. Ask “who should I use for [your service] in [your area]?” and see whether you appear. Run the same question monthly and keep a record. Server logs tell you a crawler visited. Only this tells you whether you were the answer.

If you want the budget side of this before you commit, I have broken down how much AI search optimisation costs in the UK separately.

The full checklist in one place

1. Confirm access. Google Search Console plus the ability to edit your site.
2. Check crawler access. Look for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot and Google-Extended in your logs. Check your CDN and security plugin too.
3. Fix robots.txt. Allow the search crawlers by name. Add your sitemap line. Validate before publishing.
4. One page, one job. Split any page answering more than one question.
5. Answer capsules. 40 to 60 words directly under every heading, complete on its own.
6. Headings as questions. Customer phrasing, not industry phrasing. Pull it from Search Console, People Also Ask and forums.
7. Schema. Organisation, LocalBusiness, FAQPage, Article. Validate every page. Never mark up invisible content.
8. Consistent details. Name, address, phone and description identical across Google Business Profile, Bing Places, Yell, Trustpilot, LinkedIn, Companies House and your sector directories.
9. Off-site presence. Reddit, Quora, trade press, directories. Real answers, first-person experience, regularly.
10. One original number. With sample size, date range and method.
11. Credentialled human on the page. Named byline, credentials bio, outbound citations, visible dates.
12. Test monthly. Clean-slate AI, customer’s question, keep a record.

Frequently asked questions: Optimise Your Website for AI Search

Do I need a developer to optimise my website for AI search?

For most of this, no. Steps 3 to 5 and 8 to 10 are content work you can do yourself in your CMS. Steps 2 and 6 involve editing robots.txt and adding JSON-LD, which most SEO plugins handle without touching code. You would want a developer if your site is custom-built with no plugin layer, if your content is rendered by JavaScript in a way crawlers cannot read, or if your CDN is blocking bots at server level and you cannot find the setting. That is usually an hour or two of their time, not a project.

Does AI search optimisation replace traditional SEO?

No, and I would be sceptical of anyone selling it that way. AI models still favour sites that are fast, indexable, well structured and genuinely useful, which is exactly what SEO has always been about. What changes is the extraction layer on top: answer capsules, schema, question-shaped headings and off-site consistency. If your technical foundations are broken, adding schema will not rescue you. Fix the foundations, then add the AI layer.

How do I check whether a crawler is being blocked on my site?

Start with yourdomain.co.uk/robots.txt and read it. Then check your CDN or security plugin, because Cloudflare, Sucuri and similar tools have their own AI bot controls that override robots.txt. Then look at your server logs or hosting analytics for the last 30 days and search for the AI user agents by name. If robots.txt looks fine but no AI crawlers appear in your logs, the block is at server or CDN level. If you are verifying a bot’s identity from a log entry, do a reverse DNS lookup rather than trusting the user-agent string, because those can be faked.

Can a small business realistically get cited by AI models, or is this only for big brands?

It is realistic for a business of one. Mine was. The experiment I ran was on my own site, with a Reddit profile that had two followers, and it worked. Budget and team size are not the deciding factors. Consistency across a handful of places, genuine first-hand expertise and content structured so a model can extract it are the deciding factors. A specialist local business with real experience often has a better shot at a specific question than a large brand writing generically about everything.

Should I block AI crawlers to protect my content?

You can, and it is a legitimate choice, but understand what you are choosing. Blocking the training crawlers (GPTBot, ClaudeBot, Google-Extended) keeps your content out of model training while leaving you eligible for citations. Blocking the search crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) removes you from AI answers entirely. Most businesses that come to me want the second thing and have accidentally done the first, or done both. Decide deliberately rather than leaving it to a plugin default.

How do I know if AI search is actually bringing me customers?

Honestly, attribution here is imperfect and I would not pretend otherwise. AI referral traffic shows up in your analytics under source names like chatgpt.com and perplexity.ai, so you can track visits. What is harder to track is the person who read your name in an AI answer, did not click, and searched for you directly a week later. The proxy I use is branded search volume in Google Search Console alongside AI referral sessions. If both climb after you start this work, something is working.

If your customers are asking AI and getting someone else’s name

You have a website. You have content. You have been doing the right things for years. And when your customer asks ChatGPT who they should use, a competitor comes back instead of you. That is a genuinely frustrating position, and it is almost never because your business is worse. It is usually because your site is not built for how the question is being asked now, and because you exist in one place online instead of ten.

Work through the ten steps above and you will fix most of it yourself. Steps 1, 2 and 6 alone will put you ahead of a lot of your competition, because most of them have not checked.

If you would rather not spend the next six months testing it, that is the work I do. No templates, no secret sauce, just this checklist built around your business rather than a generic one.

Get in touch and tell me what your customers are asking.