Google Ads Basics With AI: Search Ads for a Small Budget
Someone types "loose leaf tea sampler Hong Kong" into a search box. They are not scrolling idly. They are looking for something right now. A search ad puts your offer in front of that person at that moment, and you pay only when they click. For a small business with a small budget, that is one of the few ways to reach people who already want what you sell. It is also an easy way to waste money, because the system is built to spend your budget, and it will happily spend it on the wrong searches. This tutorial teaches you how search ads work, how to set up a small, careful campaign, how to use AI to write the ads and the keyword lists without inventing claims, how to avoid the common money traps and how to run it for a few weeks without being swallowed by the settings.
Difficulty: ★★☆☆☆ (no technical background needed)
Tools needed: Any chat AI, and an advertiser account on a search ads platform (this tutorial uses Google Ads as the example)
Last updated: November 2026
Overview
A search ad appears when someone searches for something related to what you sell. You choose the keywords you want to show for, write the ad, choose the page people land on and set a budget. When someone searches, an automatic auction decides whether your ad shows, in what position and what you pay if they click. You do not pay for people who see the ad and do not click.
Search advertising is different from social advertising in one important way: intent. On social media, you interrupt people who were doing something else. In search, people raise their hand by typing a query. Someone searching "buy loose leaf tea sampler" is far closer to buying than someone who happened to see a tea post. That is why even a small budget can work: you are paying for attention from people already in the market.
The same intent is what makes search ads easy to get wrong. Because people search for all kinds of things, a poorly targeted campaign shows your ad to searches that sound similar but are not what you sell, and you pay for those clicks. "Tea" could be tea leaves, tea sets, tea recipes, tea jobs or tea shops in another country. Most of the craft in a small account is saying clearly who you do and do not want.
Platforms also change their interfaces, names and features often, and increasingly push automated options that decide targeting and creative for you. Some are useful, and some are designed to spend more. This tutorial focuses on the principles that hold regardless of the interface, and treats automation with care.
The tutorial has seven steps. First, how a search ad works, and whether it suits your business. Second, the building blocks of a small account. Third, keywords and the searches you do not want. Fourth, writing the ad, with AI's help and without invented claims. Fifth, the landing page, tracking and budget. Sixth, the launch settings and the traps to avoid. Seventh, the first four weeks of running it. The next tutorial covers reading the results without panic, and the quest "ad test plan" lets you plan small tests on a practice business.
You will finish with a small campaign plan, a keyword list with negatives, three or four ad variations, a tracking checklist and a launch checklist.
Who This Is Useful For
What You Will Learn
What You Need
1. An offer people search for, with a landing page that matches it. The landing page tutorial in the previous module covers this.
2. A budget you can afford to learn with. It should be money you are prepared to spend with no guarantee of results.
3. Your persona, positioning and facts, if you have them.
4. A way to count the result you want: a sign-up, a purchase or a call. Conversion tracking is covered in Step 5.
5. A chat AI, and an advertiser account.
6. About three hours to plan and set up, then thirty minutes a week.
Information as of 14 November 2026. Platform interfaces, features, policies and automated options change often, and prices at auction vary by country, industry and day. This tutorial gives no prices, click costs or benchmarks, and names only features that have been long-standing. Always check the platform's current help pages and advertising policies, and set up your account from what you see on your screen. Nothing here is financial or legal advice.
Step 1 — How a search ad works, and whether it fits
The basic flow.
1. A person types a search query.
2. The platform looks for advertisers whose keywords relate to that query.
3. An auction runs in a fraction of a second. It considers your bid, the quality and relevance of your ad and landing page, and other factors.
4. If you win, your ad shows. If the person clicks, you pay a cost per click, usually less than your maximum.
5. The person lands on your page and either takes your action or leaves.
You are charged for clicks, not for showing the ad. So three things decide your result: whether the right people search, whether your ad earns the click, and whether your page turns the click into the action.
When search ads fit.
When they may not fit.
Search versus social, briefly.
| Search ads | Social ads | |
|---|---|---|
| Who sees it | People actively searching | People matching a profile, doing something else |
| Best for | Capturing existing demand | Creating awareness and interest |
| Typical risk | Paying for the wrong searches | Showing to people who do not care |
| What matters most | Keywords, relevance, landing page | Creative, audience, offer |
A good plan often uses both: social to create interest, search to capture the people who go and look. For a small budget, start with the channel closest to people ready to act.
Step 2 — The building blocks of a small account
Search ad accounts have a simple hierarchy. Keep yours small and tidy.
| Level | What it is | For a small business |
|---|---|---|
| Account | Your advertiser account, billing and settings | One |
| Campaign | A budget, locations, languages and a goal | Start with one campaign |
| Ad group | A set of closely related keywords with the ads for them | Two or three ad groups, each on one theme |
| Keywords | The searches you want to show for | Five to ten per ad group |
| Ads | The text people see | Two or three per ad group |
| Landing page | Where the click goes | One page per theme, matching the ad |
Organise by theme. The key rule: each ad group covers one tight theme, so the ad can match the search. For the tea shop:
Do not mix themes. An ad about "tea samplers" shown for "how to brew oolong" matches poorly and wastes money.
Set the location and language precisely. Target the places you can serve. A shop that ships only within Hong Kong should not show ads in other countries. Choose the language your customers search in. If you serve both Chinese and English speakers, consider separate campaigns, since keywords and ad text differ.
Choose the goal honestly. The platform will ask for a campaign goal, such as sales, leads or website traffic. Choose the one that matches what you want. A goal of "traffic" optimises for clicks, not for people who buy.
Beware automation by default. Platforms offer campaign types and "recommendations" that broaden your targeting, add keywords and raise budgets automatically. Some are useful once you have data. Many are designed to spend more. Start with settings you understand, add automation only when you can judge whether it helps, and read every recommendation before accepting. A high "optimisation score" is not your goal. Customers are.
Step 3 — Keywords, match types and negatives
Keywords tell the platform which searches to show your ad for. This is where small accounts win or lose.
Start with what customers actually type. Use your persona's words and your reviews. Add the platform's keyword research tool as a source of ideas, but only keep what fits your offer.
Use AI to brainstorm, then filter.
Role: a careful search advertiser.
Task: For a small online loose-leaf tea shop that ships within Hong Kong and sells a HK$198 three-tea sample box, suggest 30 search phrases a customer might type when ready to buy, grouped into three themes. Mark each phrase as: ready to buy, comparing options, or just learning. Then list 30 phrases that sound similar but would NOT be a good customer (for example jobs, recipes, wholesale, other countries, free), which I will use as negative keywords.
Do not include brand names of competitors.AI offers ideas, but it is guessing. Check them against real data later, in the search terms report.
Intent matters. Group keywords by how close the person is to buying.
| Intent | Example search | Fit |
|---|---|---|
| Ready to buy | "buy loose leaf tea sampler hong kong" | Best. Bid here first |
| Comparing | "best loose leaf tea sampler" | Good, but more competition and less certain |
| Learning | "how to brew oolong tea" | Often not ready to buy. Better for content or a free guide |
| Unrelated | "tea jobs", "tea set wholesale" | Exclude |
With a small budget, favour the first row and consider the second.
Understand match types. They control how closely a search must match your keyword. Names and exact behaviour have changed over time, so check the current definitions, but the idea is stable:
A sensible start for a small account is a mix of phrase and exact for your best keywords, avoiding broad until you have clean data and a good negative list.
Build a negative keyword list from day one. Negative keywords tell the platform which searches not to show for. They protect your budget.
Start with 20 to 50 negatives, and add more every week from the search terms report (Step 7). Many small accounts waste half their budget on searches they never wanted.
Keep each ad group tight. Five to ten keywords on one theme beats a hundred keywords everywhere.
Step 4 — Write the ad, with AI and without invented claims
A search ad is small: usually a few headlines and a couple of descriptions, with strict length limits. Because it is short, every word has to work, and because ads are regulated, every claim must be true.
What a good search ad does.
1. Matches the search. Echo the words the person typed. If they searched "loose leaf tea sampler", a headline that says so feels relevant.
2. States the benefit. Why this, not the other result? One clear difference.
3. Gives a reason to believe. A fact, a number or a guarantee that you can back up.
4. Calls to action. A clear next step: "Order the sample box".
5. Fits the landing page. The same promise, the same price, the same words.
Use your positioning and brief. The single message and proof you wrote earlier are the raw material. A headline might be the message, another might be the proof, a third the offer. Keep the voice card in the prompt.
Draft with AI, from your facts. Many search ads let you provide several headlines and descriptions that the platform mixes. Give AI the limits and ask for variations.
Role: a search ad copywriter who writes in the voice card below.
Voice card: [paste]
Task: Write 8 ad headlines (maximum 30 characters each) and 4 descriptions (maximum 90 characters each) for the keyword theme "loose leaf tea sampler" for the business below. Each headline should work on its own and make sense next to any other. Include: the keyword theme, the single message, one proof point, and a clear call to action.
Rules: Use ONLY the facts below. No invented numbers, ratings, discounts, awards, guarantees or "best/#1/leading" claims. No fake urgency. Do not mention competitors. No exclamation marks. Plain words.
Single message: [paste]
Facts: [paste]Then check the character counts yourself. AI often miscounts. Edit any headline that is too long.
Follow the platform's advertising policies. Rules typically restrict misleading or unverifiable claims, inappropriate use of trademarks, excessive punctuation and capitals, and certain categories such as health, finance and gambling. Read the current policy before you write. Ads that break the rules are rejected, or the account can be restricted. When unsure, remove the claim.
Avoid common ad mistakes.
Add the extras honestly. Platforms let you add extra information such as links to specific pages, short callout lines and a phone number. Use them with true, specific details: "Free delivery over HK$250", "Ships in 2 days", "Brew note in every box".
Write two or three ads per ad group, with different angles, such as the benefit, the proof and the offer. This lets you learn which works, as the next tutorial explains.
Step 5 — The landing page, tracking and budget
Three things must be in place before you spend.
1. The landing page. Send people to a page that matches the ad and has one clear action. Use the checklist from the previous tutorial: message match, a benefit headline, proof, honest answers to doubts and a single button. A good ad sending clicks to a weak page is the most common way small budgets are wasted.
2. Conversion tracking. You must be able to count the action you want: a purchase, a sign-up, a booking, a call. Without it, you only see clicks, and clicks do not pay the bills.
3. The budget. Decide in advance what you can afford to learn with.
Choose a bidding approach you understand. Platforms offer manual and automated bidding strategies. With little data, automated strategies that optimise for conversions may struggle, since they need data. A simple approach is to start with settings that give you control, watch the results, and move to automation when you have enough conversions. Read the current description of each option before choosing.
Step 6 — Launch settings and the traps to avoid
Before you press launch, run this checklist. Many small accounts lose money in the first week on settings.
Location and language. Only the places you can serve. Check the setting about people "in" a location versus people "interested in" it, so you do not show to people elsewhere.
Schedule. If your business cannot respond at certain times, consider limiting when ads show. A phone-call campaign at 3 am wastes money.
Devices. Check how your page looks on mobile. If it is poor on phones, fix it before spending.
Networks and extras. Platforms may default to showing ads on partner sites, in addition to search results, or to expanding to other networks. For a small account, start with search results only, and add other placements later when you can judge them.
Keyword expansion. Turn off automatic broadening you did not ask for. Review any "recommendations" that add keywords or raise budgets.
Negative keywords. Loaded before launch.
Ad approval. Check that ads are approved, and read any policy warnings.
Tracking. Tested end to end.
Billing and limits. Payment set up, budget and total cap written down.
Common traps.
Step 7 — The first four weeks
Once you launch, a simple routine beats constant fiddling.
Week 1: watch, do not tweak. Check daily that ads are approved, showing and spending as expected, and that tracking works. Make only emergency changes: a broken page, wrong location, a policy rejection. Resist changing bids or ads on day two.
Weekly routine, thirty minutes.
1. Read the search terms report. This shows the actual searches that triggered your ads and what you paid for them. Add irrelevant ones as negatives. Add genuinely good ones as keywords.
2. Check the numbers. Impressions, clicks, cost, conversions, and the ratios the next tutorial explains. Note them in a table.
3. Look at each ad. Which headlines get more clicks? Pause the clearly weakest only after you have enough data to judge.
4. Check the landing page and tracking. Still working? Still matching?
5. Write a note. What you saw, what you changed and why.
Use AI to analyse, carefully. Paste a summary table of your numbers, with no personal data, and ask for patterns and hypotheses, not verdicts.
Role: a careful search advertising analyst.
Task: Below is a weekly summary of my small search campaign. Point out (1) anything that looks unusual, (2) three hypotheses for why some ad groups perform differently, and (3) what additional information I would need to decide. Do not tell me to increase the budget. Do not make claims beyond the data. Remind me where the sample is too small.
[paste the table]Then check any arithmetic yourself.
Change one thing at a time. Adjust a keyword set, an ad or a bid, record the date and watch. If you change three things, you cannot tell which helped.
Decide with rules set in advance. The ad test plan quest in this track shows how: a metric, a number and an action. For example: "After 100 clicks, if cost per sign-up is above HK$150, pause the ad group. If below HK$80, add budget." Set the numbers from your break-even, not from hope.
At the end of four weeks, review. Total spend, total conversions, cost per conversion, which keywords and ads worked and what the search terms taught you. Compare with your break-even and your stop number. Then decide: continue, adjust, or stop. Stopping is a legitimate result. A clear "this does not pay back" has saved you money.
Keep it honest. Report what happened, including the money that bought nothing. Do not attribute every sale to ads, and do not ignore a sale because it was not tracked. Look at your own sales records alongside the platform's.
Worked Example: a tea shop's first search campaign
Mount and Leaf wants 40 new subscribers in six weeks. The owner has HK$3,000 for ads, which she splits: HK$1,200 for a four-week search test, and the rest kept for social tests. Her landing page already matches the offer: the HK$198 sample box.
Structure. One campaign, limited to Hong Kong and Chinese and English searches handled as two ad groups. She plans two themes: "loose leaf tea sampler" and "oolong tea".
Keywords. She asks AI for ideas and filters them to 12 phrase and exact keywords she would be happy to pay for. She also builds a negative list of 40 phrases: free, jobs, wholesale, recipe, course, other cities and tea sets.
Ads. She writes three ads per group from her single message and proof: "3 Teas, 3 Brew Notes, HK$198", "Loose Leaf Tea Made Easy" and "Order the Tea Sample Box". She checks every headline length and removes an AI-written "best loose leaf tea in Hong Kong".
Tracking and budget. She tests that an order shows up as a conversion. She works out her break-even: the sample box brings roughly HK$80 after product, packing and delivery, so she decides she can spend up to about HK$100 to win a subscriber, because many go on to reorder, a guess she will revisit as data arrives. She sets a total test cap of HK$1,200.
The first weeks. In week one she changes nothing. In the search terms report, she finds people searching "free tea samples" and "tea shop jobs hong kong". She adds both as negatives. By week two, one ad group is getting most clicks. By week three, she has 140 clicks, 9 sign-ups and cost per sign-up around HK$130.
The review. Her rule said stop an ad group above HK$150 after 100 clicks, and keep below HK$80. She is in the middle, above her HK$100 target. She changes one thing, the headline of the weakest ad, and checks the landing page speed, which was slow on mobile; she fixes that. In week four the cost per sign-up falls to about HK$95. She writes a note: small numbers, so it is a hopeful clue, not proof. She continues for another four weeks at the same budget, with a plan to stop if the cost rises above her limit.
Common Mistakes to Avoid
Three habits waste the most budget.
Mistake #1: Broad targeting and no negatives. The platform shows your ad for loose matches and you pay for clicks from jobs, recipes and freebie seekers. Fix: tight themes, phrase and exact keywords first, a negative list from day one, and a weekly review of the search terms report.
Mistake #2: Spending before the basics are in place. Clicks go to a weak page, and nothing is tracked, so you cannot learn. Fix: build a matching landing page, test conversion tracking end to end, know your break-even and set a total budget and a stop date before you start.
Mistake #3: Trusting automation and tinkering constantly. Accepting every recommendation spends more without clear benefit, and changing things daily means you learn nothing. Fix: start with settings you understand, read every recommendation before accepting, change one thing at a time and judge with rules set in advance.
Going Further
Build a negative keyword library. Keep a list from every account you run. It becomes a starter list for the next one.
Test ad angles. Once you have a few weeks of data, compare ads that lead with the benefit, the proof and the offer, one change at a time.
Add search to your funnel. Use your best-performing search terms as ideas for content and for your email sequences. People are telling you what they want.
What to learn next. The next tutorial explains how to read ad results without panic: what impressions, clicks, CTR, CPC, conversions and CPA mean, how much data is enough and how to decide to stop, keep or scale. It uses the same numbers you will see in your first weeks.
Key Takeaways
In this tutorial you learned:
Before you move on, a challenge. Plan one campaign on paper: a goal, two themes, ten keywords, thirty negatives, two ads for one theme, a landing page, a tracking test, your break-even number and a total budget with a stop date. Check every claim in the ads against your facts, and have someone review the plan before you spend anything.