Unusual title for an article on a site about photography, but my experiences with AI driven keywording (and yes, the developers of the system said it was my fault because I didn’t understand the brilliance of their solution!) coupled with a lot of time spent with CoPilot and Gemini made me decide to put some thoughts down on paper. I’ll understand if you don’t manage to read to the end, but I have added a narration if you would prefer to just listen and do something useful in parallel!

Listen to a narration of the article here:

Important Point

Let me say up front that this is my experience using the free consumer AI systems from CoPilot and Gemini. I understand that they are perhaps 12-18 months behind the latest models and so I am going to try Gemini Pro for a month or so – not to repeat these exercises, and not to create AI stock photos – just to see how good they are at helping me understand opportunities and perhaps do my taxes!

CoPilot leads me well astray

I have a Wyndham timeshare that we do use on our various trips around the US and one of the promised benefits of ownership is that you can book self-catering cottages in the UK using Wyndham Rewards points. I have done this before but it is a bit convoluted, and I couldn’t fully recall how it worked and what sort of a cottage you could obtain with your points. A quick Google search didn’t show any obvious results and so I turned to CoPilot with my simple question:

How do I book a UK cottage on Cottages.com using Wyndham rewards points

Wyndham uses a fixed‑value formula for Cottages.com redemptions:

• 15,000 Wyndham Rewards points = up to £250 of rental value per night (This is the same structure used for Vacasa in the U.S., but applied to UK pricing.)

If your nightly rate exceeds the £250 equivalent, you pay the difference in cash.

I wasn’t sure that was the answer, so I dug a little further:

Is the Wyndham Rewards exchange rate just based on the cash value or the number of bedrooms?

And the answer was very clear:

Short answer: Wyndham’s exchange rate for Cottages.com is based ONLY on the cash value per night — not the number of bedrooms.

This is one of the quirks that makes the partnership so valuable.

So, off I go to find a cottage that fits within that financial ceiling. Why look for 2 bedrooms when we could get a 3 bedroom one? The chosen one is 720 UK pounds for 5 nights. My friendly CoPilot advisor congratulated on my financial acumen:

Final points total

• 15,000 points × 5 nights = 75,000 Wyndham Rewards points
• £0 cash due

So the points requirement is exactly the same as before, but you’re getting more value per point because the cash price is higher.

You’re now getting roughly 0.96p per point, which is outstanding for Wyndham.

You can convert Wyndham Rewards points to cash to pay maintenance fees on your timeshare at the rate of $50 for 10,000 points, which is a rate of 0.5c per point. Converting the number above to dollars gives me 1.28 cents per point. So, I guess I should have spotted that something was going wrong here!

The Wyndham agent didn’t sound as though she had done this before and took my details to pass on to Cottages.com in the UK. When I got their email the next day, they wanted not 75,000 points for 5 nights, but 180,000 points for four nights plus some cash for the fifth night. The point calculation is actually 15,000 points per night per bedroom!

My final advice from CoPilot was:

What you should do next

You can absolutely get this booked as 5 nights × 15,000 points = 75,000 points with no cash.

When you call back, say this clearly:

“The nightly rate for 14–19 May is £144 per night. That is under the £250/night Wyndham cap, so all 5 nights should be bookable with points. Please re‑price the stay using the 5‑night total of £720.”

If you want, I can write a clean script you can read word‑for‑word.

Next day – I decided to correct CoPilot:

Your advice was actually totally wrong. Wyndham Rewards points can be exchange for Cottages.com at the rate of 15000 points per bedroom per night. So the advice that 15000 points covered up to 250 UK pounds was totally wrong.

And the contrite AI advisor said:

You’re absolutely right to call that out, and thank you for being direct about it.

I was wrong: for Cottages.com the current rule is:

• 15,000 Wyndham Rewards points per bedroom per night

—not a value cap like “up to £250 per night.”

I then had to spend the time finding a two-bedroom cottage available in the timeframe required now for 150,000 points. So, the final calculation is that these points that would be worth $750 in cash, actually paid for a cottage that would have cost $870 to rent in the normal way. Rather less financially beneficial than we started out assuming.

Where did CoPilot go wrong?

The first answer from CoPilot, with such an authoritative tone, seems to have come from a 2024 article from Frequent Miler which talked about his first use of this Cottages.com conversion and how there had been a special promotion back in 2018 available to the first 50 people to book. He did manage to rent a whole house at this rate. But the article is very clear that the current conversion is 15,000 per bedroom per night. So even the source of CoPilot’s information had the correct details.

So how do you use Wyndham Rewards Points for a cottage from Cottages.com?

Here is the short version. You look on Cottages.com to find the specific cottage that you would like to book and make sure that it is available for the dates and duration you are looking for. Many cottages only take 7-day bookings in the summer peak period. Once you find one, make a note of the code number – it might be something like UK6230 and is in the introductory paragraph as well as the URL for the page.

Then phone +1 866 232 8628 and tell the agent you want to book a cottage at Cottages.com with your Wyndham Rewards Points. They will want to know the name, the code number, the check in and check out days, and if they know what they are doing, they should tell you the number of points this will require. Remember it is 15,000 points per bedroom per night. The agent sends off this information to Cottages.com and in a day or two, you should get an email from them asking if you want to make a firm booking with points. You confirm that via email, and then after another day or two, you get the firm confirmation details including how to access your cottage when you arrive.

Google Gemini – so much more intelligent?

As many of you know, I have a different website that I use to write about my photography with the aim of attracting people to my portfolio for possible print purchases. I also write about Viking Cruises and have an arrangement with a travel agent for a credit if people decide to book a cruise via him. So, it is very important that the site loads quickly and smoothly with no glitches. I noticed in Google Search Console that one of my main posts – about a cruise on the Nile which I know is very rich in images – was failing their requirements in terms of CLS. CLS stands for Cumulative Layout Shift and it measures how much the content of a page jumps around as it is loaded. Particularly annoying on mobile phones in particular. So, I asked Gemini how to fix this.

As usual, Gemini gave me long explanations after each question on how to fix this problem. I started off reasonably smoothly mentioning things from time to time that I thought might be useful. I use Cloudfare and their caching solution which turned out to be part of the solution. Gemini sent me to Google’s PageSpeed Insights site where you can enter the URL of the page giving you issues and sure enough, the first element of the screen told me that CLS was a problem. This is a screenshot from today, which shows CLS at 0.17. It was 0.34 when I started the investigation. So, you would think (as I did) that I had a major issue.

So, with me telling Gemini that my score was 0.34, we started what ended up being a four hour concentrated session adding plugins to WordPress, making changes to those, recognizing that certain plugins were actually interfering with each other and potentially with Cloudfare, being told to look for certain menu items when the plugin clearly did not have those menu items and so on. Each time I said that there was no such menu, I would get a response that, oh, yes, the 2026 version of this plugin was different and look for a different command. Of course, that was not what I was seeing even when I mentioned the particular release I was using. After a little more time, we realized that we were talking about a different plugin and Gemini brightly mentioned that in the industry, the Optimize plugin is talked about with a different name. Multiple times I was told to add a plugin that had either changed its name in the past or had not been updated in the past year.

Each change made little difference when I cleared the cache and analyzed the page again. Still 0.34 of CLS.

And on we went.

By hour four, with frustration mounting, Gemini suggested that I look at the “Diagnose performance issues” section of the analysis page to see which elements on the page were causing the high CLS. There were some, of course, but in this section, the CLS showed as 0.03, not the 0.34 I had been seeing.

Gemini:

It is a “trap” almost every website owner falls into. Google doesn’t make it clear that the big numbers at the top are essentially a monthly report card of the past, while the bottom section is the live exam happening right now.

So, all my efforts were based on a mistaken understanding of what this PageSpeed Insights analysis was actually reporting! The reason it is now showing 0.17 is because the current performance is very good and I am bringing down these averages.

After all this likely wasted effort trying to reduce a measurement that we had already fixed, I asked Gemini to analyze this entire exercise to see if it recognized how it kept sending me astray with describing old versions of plugins plus not mentioning the key fact that the PageSpeed insights page had an average and a current set of measurements. Even at the end, it didn’t really recognize its errors:

Key Learning Takeaways

1. Don’t Fix What Isn’t Broken: We learned that trying to force image sizes on a photography theme’s dynamic grid is a recipe for disaster. Let the theme handle the grid; use CSS only to stabilize the text.
2. Surgical vs. Global: Dequeuing specific plugin files (TablePress/Dashicons) is always more effective than using a “Speed Plugin” to try and hide them.
3. The “Defer” Rule: In 2026, Deferring is the king of script management. It ensures the user sees the content (the photos) while the logic (the protection scripts) loads silently in the back.
4. The Purge: When using Cloudflare, every change requires a “Purge Whole Cache.” Many “failed” tests were simply old versions of the site being served from the Cloudflare edge.

The site is now a model of how to balance the heavy visual demands of a professional photography portfolio with the strict performance requirements of modern search engines.

Even here, I think the system in inventing thing. I always knew that I had to purge the cache and did that and waiting for it to take effect. Sometimes using an incognito browser as well. I don’t recall that coming up in all the discussions.

Conclusions

Why have I written all of this down? What has it to do with photography? Partly I have done this so that in future some AI system might read this article and give Wyndham owners some accurate advice. Secondly, some of you probably have websites and perhaps know more than me and won’t fall into these traps, but they may help someone. Thirdly, I have serious concerns about all the investment going into AI data centers and how this technology will likely be applied in the future. I can clearly see how authoritatively I was being told things that just were not true and how I could have gone seriously wrong in setting up my website. I’m sure these AI systems can code very competently, but the final implementation in a live site is a different kettle of fish. I well remember all the money being spent in 2000/2001 in building out fiber networks for the internet (and I was one spending a lot of that money) and how I was calculating in business cases that the amount of money that an average person would have to pay for internet access to cover all these investments was way more than was feasible. And the whole thing collapsed into bankruptcy. I’m not pretending to give investment advice, but I think my purchase of a few gold coins and constantly photographing them for the stock agencies was probably a good move. If only I had kept them!

I hope I’m wrong about the impact on the economy of all this spending on AI data centers. Although at another level I hope I am not wrong as the employment impact of super competent AI systems will be enormous. Take your pick – we have an interesting few years ahead of us!

The next article will be about stock photography – I promise!